‘This technology may be scaled to monitor the well-being of elderly people or just identify suspicious behaviours at home,’ scientists claim

Scientists have figured out how to identify people in a building by using artificial intelligence to analyse WiFi signals.

A team at Carnegie Mellon University developed a deep neural network to digitally map human bodies when in the presence of WiFi signals.

The researchers said they developed the technology in an effort to address the limitations of current 2D and 3D computer vision tools like cameras, Lidar and radars.

A pre-print study of the technology describes how the system “can serve as a ubiquitous substitute” for these other methods.

“The results of the study reveal that our model can estimate the dense pose of multiple subjects, with comparable performance to image-based approaches, by utilising WiFi signals as the only input,” the researchers wrote in the study.

“This paves the way for low-cost, broadly accessible, and privacy-preserving algorithms for human sensing.”

The researchers used a system called DensePose, which was developed by AI researchers at Meta’s Facebook, to measure the phase and amplitude signals emitted from WiFi routers.

The WiFi-based model is currently limited to 2D models of people, however the researchers hope to keep developing it in order to be able to predict 3D human shapes.

The team claims the device offers a cheap and “privacy-friendly” alternative to cameras and Lidars, as it does not provide a clear image of the subject.Potential applications include assisting elderly people, the scientists noted, allowing them to continue to live independently.

“We believe that WiFi signals can serve as a ubiquitous substitute for RGB [camera] images for human sensing in certain instances,” they wrote.

Scientists find way to photograph people through walls using WiFi

The technology can be used to map out the inside of a building in 20 to 30 seconds

Scientist at the Technical University of Munich have come up with a new technology that can take 3D photographs of objects inside a room or building using WiFi.

The research, which was published by Physical Review Letters, shows that wireless data transmission systems including WiFi or Bluetooth emit electromagnetic waves with a precisely known amplitude and phase which can be used to map objects in their surroundings.

“Propagating in space, this radiation forms a hologram — a two-dimensional wave front encoding a three-dimensional view of all objects traversed by the light beam. Three-dimensional views of objects and emitters are made by feeding the resulting data into digital reconstruction algorithms,” Philip Holls and Friedemann Reinhard noted in their report.

PHOTO: Philip Holls and Friedemann Reinhard/Physical Review Letters

The two scientist went on to test their ground breaking technology in an experimental environment. Stray radiation from a commercial router was employed to image metre-sized objects by digital holography. The researchers then record a hologram of WiFi radiation by using a synthetic aperture approach.

The scientist then employ digital implementation of dark-field propagation to suppress multipath reflection, which significantly enhances the quality of the resulting images.

PHOTO: Philip Holls and Friedemann Reinhard/Physical Review Letters

“It can basically scan a room with someone’s WiFi transmission,” said Holl, a 23-year-old undergraduate physics student at the Technical University of Munich. He initially built the device as part of his bachelor thesis with the help of his academic supervisor, Friedemann Reinhard.

“If there’s a cup of coffee on a table, you may see something is there, but you couldn’t see the shape,” said Holl. “But you could make out the shape of a person, or a dog on a couch. Really any object that’s more than four centimetres in size.”

PHOTO: Philip Holls and Friedemann Reinhard/Physical Review Letters

The technology, which is at its early stages, has application in a wide range of areas. Adding an array of reference antennas, said Holl, inside a truck, might help rescue workers detect people in rubble left by an earthquake or spy agencies see whether anyone is home.

“You could probably use a drone to map out the inside of an entire building in 20 to 30 seconds,” he said. source

MIT’s AI can now ‘see’ and track people through walls using wireless signals

The creepy RF-Pose system developed by MIT’s CSAIL can track a person through walls or identify one specific person out of a group of 100 people.

 

Credit: Jason Dorfman/MIT CSAIL

Scientists Can Now Use WiFi to See Through People’s Walls

• Researchers at Carnegie Mellon University can map human bodies through walls using WiFi signals.
• The technology tracks key points on the body for detection, extending previous research into using WiFi signals to locate humans.
• The study’s authors position the breakthrough as helpful to privacy, even if it opens the possibilities for much easier, cheaper human tracking.

It isn’t immediately clear how using only a WiFi signal to track human movement through walls improves personal privacy … but that’s what a new study from Carnegie Mellon University claims.

In a recently published paper, the researchers expanded on the study of employing WiFi signals to map human movement, especially in low-light situations that make using other technologies less than desirable. They write:

“We developed a deep neural network that maps the phase and amplitude of WiFi signals to UV coordinates within 24 human regions. The results of the study reveal that our model can estimate the dense pose of multiple subjects, with comparable performance to image-based approaches, by utilizing WiFi signals as the only input.”

This pathway opens the options for low-cost, broadly accessible human tracking through walls.

High-cost technology has successfully mapped people’s movements through walls for years. And researchers at the Massachusetts Institute of Technology have spent over a decade working on ways to more easily see people through walls, whether using cell phone signals or WiFi. In the Carnegie Mellon study, scientists had WiFi signals send and receive a body’s coordinates and then used DensePose to map the body.

From the study:

“Advances in computer vision and machine learning techniques have led to significant development in 2D and 3D human pose estimation from RGB cameras, LiDAR, and radars. However, human pose estimation from images is adversely affected by occlusion and lighting, which are common in many scenarios of interest.”

By reducing the need for the advanced—and expensive—technology, the Carnegie Mellon researchers say they can make human tracking more available. Somehow, they’ve also positioned the breakthrough as a privacy-positive situation. You don’t need WiFi signals to see that’s a bit sketchy. source

AI Learns How to Use Wi-Fi Signals to See People Through Walls

Scientists have found a way to use artificial intelligence and WiFi signals to recognize people in a building, just like James Bond.

A team at Carnegie Mellon University created a deep neural network that can map human bodies using WiFi signals.

The researchers developed this technology to overcome the limitations of current 2D and 3D computer vision tools.

In a study, they explained that their model can accurately estimate the positions of multiple people using only WiFi signals.

They hope that this breakthrough will lead to affordable, widely available, and privacy-friendly algorithms for detecting humans.

AI can ‘see’ people through walls using WiFi signals

Scientists have figured out how to identify people in a building by using artificial intelligence to analyse WiFi signals.

A team at Carnegie Mellon University developed a deep neural network to digitally map human bodies when in the presence of WiFi signals.

The researchers said they developed the technology in an effort to address the limitations of current 2D and 3D computer vision tools like cameras, Lidar and radars.

A pre-print study of the technology describes how the system “can serve as a ubiquitous substitute” for these other methods.

“The results of the study reveal that our model can estimate the dense pose of multiple subjects, with comparable performance to image-based approaches, by utilising WiFi signals as the only input,” the researchers wrote in the study.

“This paves the way for low-cost, broadly accessible, and privacy-preserving algorithms for human sensing.”

The researchers used a system called DensePose, which was developed by AI researchers at Meta’s Facebook, to measure the phase and amplitude signals emitted from WiFi routers.

The WiFi-based model is currently limited to 2D models of people, however the researchers hope to keep developing it in order to be able to predict 3D human shapes.

The team claims the device offers a cheap and “privacy-friendly” alternative to cameras and Lidars, as it does not provide a clear image of the subject.

Potential applications include assisting elderly people, the scientists noted, allowing them to continue to live independently.

“We believe that WiFi signals can serve as a ubiquitous substitute for RGB [camera] images for human sensing in certain instances,” they wrote.

“Most households in developed countries already have WiFi at home, and this technology may be scaled to monitor the well-being of elderly people or just identify suspicious behaviours at home.” source

This neural network uses WiFi to see through walls

“Yes, a simple WiFi signal can be used to spy on us through walls, a scary thought for people who value their privacy”, Nigel elaborates.

Seeing through walls is an ability most people associate with Superman’s X-ray vision or James Bond’s X-ray sunglasses but scientists at Carnegie Mellon University have come up with a way to see through walls using a neural network and WiFi signals. While previous efforts by MIT researchers used radio waves to see through walls, the images were typically low resolution and difficult to use to identify people.

WiFi signals, however, are apparently even better than regular RGB cameras for detecting humans and can be hooked up to a neural network in order to create a 3D map of a human body standing on the other side of a wall. Yes, a simple WiFi signal can be used to spy on us through walls, a scary thought for people who value their privacy.

WiFi Vision

As opposed to a camera that works on the principles of light and visibility, this innovation sends a low-powered WiFi signal through a wall in order to detect everything in the room on the other side of it. It does this much like a radar signal that bounces off static objects to detect movement. In the case of human bodies, the WiFi signal sends the coordinates of key points on the human body like the head, arms, and legs (that determine posture) which are then fed into a system called DensePose which assimilates them into a 3D picture. DensePose is a deep neural network that can map all the pixels from a 2D RGB image of the human body and convert them to a 3D image.

Developed by a team of researchers from London along with Facebook’s AI team, the process involves establishing a “dense” correspondence between a 2D image and a 3D surface model. In this case, however, it’s not 2D images that are being mapped but rather the results of a WiFi-powered scan that is being mapped to a 3D image.

The efficiency and accuracy of details are quite astounding as you can see in some images on the DensePose website. While this explanation could be interpreted by some to be overly simplified since there’s quite a lot of mathematics involved in this process of “pose estimation,” the short version is that the technology does work and will only require off-the-shelf WiFi equipment to use.

Security Implications

Now while researchers argue that this innovation could be progress for privacy rights since we won’t need cameras everywhere anymore, that’s sort of like making the TV remote shoot lasers and then saying the good part is we don’t need guns anymore. If every lunatic that could get their hands on a TV remote could use it to blast things out of their way, there wouldn’t be a lot of things left to blast. Similarly, if every nosey person could use a WiFi router to see through walls, it would officially be the end of privacy, can you imagine if the paparazzi got their hands on something like this? No one would be safe, especially since walls are something most people expect to be able to rely on for privacy.

Now while there is always an upside to new technology like the potential to use it for law enforcement or even home security purposes, like in the case of nuclear technology, it always cuts both ways. Researchers at Waterloo University have devised a way to use WiFi to see through walls using a store-bought drone and an additional $20 worth of store-bought equipment.

While the Wi-Peep doesn’t have a fancy 3D mapping application powering it, what it does have is a security loophole called Polite WiFi. What this means is that every WiFi-powered device, even secure, will respond to contact attempts from other devices in range. These responses are then used to track and locate the wearers of these devices or the devices themselves.

X-ray Vision

With the amount of data we generate on a daily basis, there are those who would argue that our phones already know more about us than we possibly ever could. That being said, however, the possibility of being spied on every time you operate a WiFi device might be something we may just have to get used to in the future. In addition to walls, WiFi also overcomes a number of obstacles like poor lighting, occlusion, or even the need for a camera.

There’s no argument that it would be great for the armed forces and covert military operations where knowledge of the insides of enemy buildings is generally hard to come by unless you have someone inside. If X-ray vision is a superpower we are going to have soon, we may as well enjoy it. As for privacy, for all we know, it may already be dead. source

WiFi routers can see people through walls. Here’s how it works

With the help of AI, the researchers were able to detect the movement of human bodies in a room using Wi-Fi routers — even through walls.

Researchers at Carnegie Mellon University have repurposed a common piece of tech found inn virtually every household into a tracking technology.

They used radio signals from WiFi routers to detect and track the three-dimensional shape and movements of human bodies in a room. They didn’t have to use any cameras or expensive LiDAR hardware.

“We believe that WiFi signals can serve as a ubiquitous substitute for RGB images for human sensing in certain instances. Illumination and occlusion have little effect on WiFi-based solutions used for interior monitoring. In addition, they protect individuals’ privacy and the required equipment can be bought at a reasonable price. In fact, most households in developed countries already have WiFi at home, and this technology may be scaled to monitor the well-being of elder people or just identify suspicious behaviors at home,” the authors wrote in their study, which is yet to be formally peer-reviewed and is available on the preprint server ArXiv.

AI turns WiFi routers into cameras that can see through walls

The team used DensePose, a system for mapping all of the pixels on the surface of a human body in a photo, developed by researchers at Facebook’s AI lab and a London-based team. What makes DensePose powerful is its ability to identify over two dozen key points and areas in the human body, such as joints and body parts like the arms, head, and torso. This allows the AI to describe a person’s pose.

Combining all of this data with a deep neural network, they were able to map WiFi signals’ phase and amplitude sent and received by routers to coordinates on human bodies. source

The world’s first fully autonomous AI software engineer was unveiled Tuesday, according to creator Cognition.

“Devin is an autonomous agent that solves engineering tasks through the use of its own shell, code editor, and web browser,” said the company on X.

New platform /ask Swimm is most recent development in AI software

Devin: What can it do?

Artificial Intelligence in Music is Creating More Creativity

The contribution of artificial intelligence in music is less popular. Music is regarded as one of the most human characteristics. So, what happens if it’s generated by a machine? Machine learning models are used by AI to build new correlations and patterns from the data it receives. There are presently millions upon millions of tracks accessible for streaming in the music industry. The ability of artificial intelligence in music, to learn and iterate depending on its knowledge has resulted in some significant developments. 1. Creating Music: There are 12 notes in modern music, as well as 24 major and minor scales. When you combine these factors with various tempos, harmonies, and voices, Artificial Intelligence has the ability to generate millions of tunes. Musicians may now produce music utilizing simple production, remixing, and mastering techniques thanks to artificial intelligence. These allow artists to generate endless variations of melodies and give a vast creative playground. Some popular AI software is AIVA, ChucK and etc. 2. Virtually creating pop stars: During the pandemic, people can’t go to musical concerts. Authentic Artists has released a vast collection of AI-powered virtual artists capable of creating new musical experiences. Their animated virtual musicians create all-original songs to perform on screen, and they also respond to crowd response by changing the speed or intensity, or even skipping to the next song in the set. 3. Finding future artists: Artificial intelligence aids the music industry’s A&R (artist and repertoire) discovery by sifting through songs in search of the next future artist. The application of Artificial intelligence in music production is absolutely astounding, and it will have a big influence in the near future. Our world will continue to sound nicer every year as AI software improve and the music industry learns how to utilize artificial intelligence as a supplement to human creativity. source

Music and Sample Generation

One of the most contentious areas of AI is content creation. This is where many traditional composers scoff and shake their heads in disgust. This could be a reaction to the often subpar or banal results generated by AI content creators. Or, maybe it’s the underlying fear of being replaced. But fear not humans, AI content creation comes in many forms and can be a useful tool in the right hands. In terms of banality, a quick tour of YouTube instructional videos will reveal an endless supply of terrible music created by humans, so why shouldn’t AI have a crack at it as well?

Arcade by Output

Many people are using Arcade and taking advantage of their ever-growing library of sampler instruments. But you can also use their Kit Generator to input your own samples. The software analyzes the sound and generates a unique kit complete with effects processing. You can change the way Arcade chops and slices your sound using four distinct modes. While there is no mention of AI-based procedures on the Output site, this process seems really smart to me.

Emergent Drums by Audialab

This software uses AI to create drum samples from scratch guaranteeing unique royalty-free samples. The user can alter parameters at will to modify existing sounds on a continuum from similar to completely random. The GUI is a familiar drum-pad layout with additional controls for pitch, panning and amplitude envelope.

Pilot Plugins by Mixed in Key

These plugins include Pilot Melody, Pilot Bass and Pilot Arpeggio. They will create content based on a chord progression you can edit in the plugin GUI, or you can connect it to one of their other devices called Captain Chords. It seems to be oriented toward techno, future rave and house genres. The devices are very user friendly and provide intuitive parameters such as density, syncopate and note length — each can be altered on the fly. You can export content as MIDI or audio with drag-and-drop ease.

Scaler 2 by Plugin Boutique

Scaler can “determine what key and scale you’re in and suggest chords that match your music. The plugin helps you discover the perfect melody with performance expressions, and it can even suggest ways to change from one key to another.” It has a robust collection of content that includes chord sets, phrases, sequences, basslines, melodies, rhythms, modulation pathways and chord substitutions.

ORB Producer Suite by Hexachords

This software boasts AI computing capable of an infinite number of patterns, melodies and basslines. I wrote about the beginnings of the software then called Orb Composer back in 2018, and it has come a long way since then adding polyrhythms, block chaining (“lets you create one long piece of music from all of your themes for continuous playback at the click of a button”) and more.

Amper Score by Shutterstock

Online services like Amper makes creating royalty-free music with AI brutally simple: select genre and length, press the button, tweak and download. The results are pretty stunning in terms of audio quality and the speed at which content can be generated. For non-composers that need tons original music for things like video or podcasts, this is an extremely attractive service. It will definitely reduce the need for music library content composed in the traditional way. Production music composers will need to adapt to this new arena and learn how to inject this automated process with a healthy dose of humanity.

Aiva

Here’s a track composed by Aiva, another AI composing platform:

Artist Benoît Carré suggests the next step in AI-generated music will involve delving further into deep learning techniques, while on a personal level, he hopes to develop a live show based around his developments in AI. “I want to show that there is an artist behind the project. Often one of the issues people have with artificial intelligence, is that they think that the artist is just pushing the buttons — and that’s not the case… there’s still a role for composers to be pushing that envelope forward, because at the moment AI is only capable of doing what has happened before.” (source)

The key phrase in this quote for me is that last bit, “at the moment,” because the singularity is coming and no one really knows what that will mean. source

9 Best AI Music Generators For Creative Inspiration

SOFTWARE AND SERVICES

1. AIVA

AIVA, which stands for “Artificial Intelligence Virtual Artist,” is a software platform that uses deep learning algorithms to compose original music tracks.

Unlike many other AI music generators, AIVA focuses on the music theory aspect of music creation. It analyzes and learns from a vast database of music, enabling it to extrapolate data and create original music pieces in various styles and genres.

To create music with AIVA, you input your preferences into the AI, such as the tempo, mood, genre, and other parameters, and AIVA uses its algorithms to compose a piece of music based on your input. The resulting track is separated into different instrument tracks, which can be edited or exported into a MIDI file to use with your own samples.

AIVA has a clean and intuitive interface, with features that make it suitable for both seasoned composers and those with no background in music theory or production.

It offers a full-fledged integrated editor that functions like a Digital Audio Workstation (DAW), where you can edit MIDI tracks and apply various effects to the generated track.

AIVA also features an achievement window, making it a gamified experience, and its own radio, which plays AI-generated pieces.

Pricing:

• Free (€0)

• Standard (€15/month)

• Pro (€49/month)

Read more: “AIVA – Everything You Need To Know – AI Composition“

2. Soundraw

Soundraw is an easy-to-use AI music generator that simplifies the process of creating original and royalty-free music for various creative projects.

Its interface is reminiscent of other stock audio music services, but it stands out from them by offering dozens of variations of a song that can be customized to fit specific needs.

Soundraw’s AI algorithms analyze a vast database of musical patterns and structures based on user preferences, allowing users to select the mood, genre, theme, length, tempo, and instruments to generate custom compositions.

Soundraw also allows for music editing, with a fun and simple to ease “sound blocks system” where users can remove or change the intensity of different sections of the song to design the progression that best fits their project.

It also features a Pro mode with additional depth and options for more advanced users who want further customization.

Soundraw stands out for its simplicity, being a powerful tool for content creators, music producers, video editors, podcasters, and game developers looking for efficient and customizable AI-generated music for their projects.

Pricing:

• Free (€0)

• Personal (€19.99/month)

Read more: “Soundraw Review – Easy AI Music Generation“

3. Mubert

Mubert AI is a music generation tool that also uses artificial intelligence to create custom music streams tailored to your preferences.

Just like Soundraw, Mubert AI uses a vast dataset of music and customer feedback to generate high-quality original songs and then presenting it with an audio stock service interface.

Most of the things said about Soundraw, can be said about Mubert as they are basically offering the same kind of service and coming from very similar angles, making Soundraw and Mubert the “Microsoft and Apple of AI Music Generation”.

While Soundraw offers more customization with its editing features and Pro mode, Mubert stands out by offering different products like “Mubert Studio” where you can earn money with your tracks and samples in collaboration with its AI system and “Mubert Play” where you can listen to some AI-generated music, carefully calculated my Mubert’s smart algorithms to fit your musical tastes.

It also offers a mobile app, so you can experiment and listen to AI-generated music on the go.

Pricing:

• Ambassador ($0)

• Creator ($14/month)

• Pro ($39/month)

• Business($199/month)

Also read: “Mubert Review – Everything You Need to Know“

4. Musenet (OpenAI)

From the same creators of ChatGPT, MuseNet is an exciting AI music generator that allows musicians and non-musicians alike to compose music.

It functions with a simple interface that features a MIDI grid, and has two modes: simple and advanced.

In simple mode, you can choose a composer or style, an optional start of a famous piece, and start generating random samples.

In advanced mode, you can interact with the model directly and create an entirely new piece, although it could take a bit longer for the completions.

One of the most unique features of MuseNet is its ability to generate music in the style of various composers and bands, such as Chopin for example.

It also has a visual interface that shows how different composers and styles relate to each other using a 2-D map.

MuseNet has some limitations, such as the instruments you ask for being interpreted as suggestions rather than requirements, and it may have difficulty with odd pairings of styles and instruments.

While it used to be previously available to the public and available to use, it is (at the time of writing this article) not available, but still deserves a spot on this list.

Pricing:

• Currently not available

5. Riffusion

Riffusion is a groundbreaking neural network developed by Seth Forsgren and Hayk Martiros that generates music using images of sound instead of audio.

It is basically a fine-tuned version of Stable Diffusion, a very recognized open-source model for generating images from text prompts, but applied to spectrograms. Riffusion uses the same principles of image generation from Stable Diffusion, but focused on interpolation and looping of images. The generated images are then processed through a process called “Short-time Fourier transform” to extrapolate the information into audio waves, resulting in music.

While technical jargon sounds a little bit intimidating, the tool itself is ridiculously easy to use.

You don’t have to create any account to try it, and all you need to is literally open the website and start prompting right away.

You can also use the random prompt button for suggestions. The interface is very simple and straightforward, with options to play, share, and adjust settings such as seed image and denoising.

Pricing:

• 100% Free-to-use

Read more: “Riffusion – The Stable Diffusion of Music“

6. MusicLM

MusicLM is an AI-powered music generator tool developed by another known: Google

Just like many of the other AI music generators on this list it can turn text descriptions into music, but it stands out by going a step further and letting the user whistle or hum a melody for the AI to follow.

MusicLM works by analyzing the text for musical attributes such as tempo, melody, rhythm, and harmony, and uses a hierarchical sequence-to-sequence modeling technique to generate the music

This model is already capable of many interesting things, like for example “Painting Caption Conditioning” where it turns descriptions of paintings into audio interpretations, potentially allowing us to hear images.

MusicLM is still in research phase, but there is already a research paper available online with some samples available to listen.

Pricing:

• Currently not available

Read more: “MusicLM: Google’s Promising AI For Music Generation“

7. Beatoven

With its clever name combining (“Beethoven” + “Beat”), Beatoven is another great AI music generator that promises to simplify the process of music creation.

One of the coolest things about Beatoven, is the fact it is trained through the contributions of many real artists, making it one of the most transparent and ethical AI music creators in the market right now.

Beatoven offers 8 different genres to choose from as well as 16 different moods, so you can easily match the music to the tone and theme of your content.

It then allows you to make cuts and adjustments, giving a familiar feel to other simple music editors, like Audacity and Ableton.

Whether you need an upbeat track for a travel vlog, a soothing melody for a meditation podcast, or a catchy tune for a promotional video, Beatoven.ai has enough features to fill those needs.

Pricing:

• Basic (Free)

• Pro($20/month) for 2 users

• Discounted prices for more users

8. Melobytes

Melobytes is more than an AI Music Generator. Melobytes is a huge Artificial Intelligence playground where you can experiment using AI in dozens of different ways.

This website features tons of different apps you can play around with, from random story generators and crossword solvers powered by AI to different kinds of music generators with different functions.

Melobytes features a whole section just for music composition with tools like: “Drawing to Music”, “Movie Music Maker” and “Text to rap song”, just to name a few.

It also has a MIDI section capable of making different file conversions using MIDI and a section where you can create Algorithmic Music and lyrics.

It has so many features that it could warrant a 5000 words article just to scratch the surface, so we urge everyone to try and explore this amazing tool by themselves, while we work on that article.

Pricing:

• Free (€0)

• 7-day Pass (€9.90)

• 1 Month (€13.90)

• 1 Year (€49.90)

9. Ecrett

Just like Soundraw and Mubert, Ecrett Music is a cutting-edge AI-powered music generator that allows users to create unique royalty-free music, soundscapes, and sound effects.

It has an intuitive interface and a vast library of sounds, instruments, and effects, and like Soundraw, it features an easy-to-use editor that is perfect for musicians, sound designers, and creative individuals of all skill levels.

Easily mix and match different elements to craft your own music, and search for sounds and instruments by genre, mood, or key.

It also one of the cheapest solutions in the market, making it an easy choice for any content creator that needs royalty free music without breaking the bank.

Pricing:

• Free ($0)

• Individual ($4.99)

• Business ( $14.99)

source

AI & Music

What Does Artificial Intelligence Mean For Musicians & Producers?

Does AI threaten to put us all out of our jobs — or will it simply make us better at doing them? We talked to the developers leading the AI revolution in music.

Artificial intelligence is making an impact almost everywhere. Self-driving cars and virtual assistants are promising to revolutionise everyday life, and music production is also a fertile area for research. So what does AI have to offer music producers, and is it a boon or a threat to people who make music for a living? To find out, we interviewed some leading experts in AI music production. Their responses offer a fascinating insight into how current research and developments could change how we produce and listen to music in the future.

What Is Artificial Intelligence?

Conventional computer software such as a digital audio workstation or plug-ins consists of a set of instructions created by a computer programmer. These instructions interpret the data and user input we provide when interacting with the software, and the software calculates a set result. For example, when we use a compressor plug-in to process a vocal recording, the plug-in responds to our settings and to the dynamics of the material in a predictable way. This can be very useful, but its effectiveness depends entirely on how we set up the compression parameters. A different vocal recording might require a different compression ratio, or indeed a completely different processing chain. Beyond providing presets that we can try out, conventional software can’t help us make this decision. Instead, we draw upon our own past experiences and learned skills to decide the settings for the new vocal recording.

Potentially, an AI compressor algorithm could analyse a particular vocalist’s performance, compare it with a library of other performances and generate appropriate compression settings.Artificial intelligence software could emulate these cognitive skills. Rather than simply offering a dial for the user to set compression ratio, an AI-based compressor could figure out for itself what the best setting might be. Through the analysis of large numbers of existing vocal recordings, the software would build up a mathematical model of what ‘good vocal compression’ sounds like, and apply this model to new recordings by comparing them with the patterns it has learned, this setting the compressor on the basis of its own learned experience, rather than ours.

In our scenario, the user simply wants the vocal to be more consistent in level. AI software could apply its model learned from earlier vocal recordings to automatically adjust the compression ratio, threshold and other settings, without requiring the user to apply his or her own past experience, learned skills or decision-making.

This is a modest theoretical illustration of how computing concepts such as artificial intelligence, machine learning and big data could offer new possibilities. These computing concepts are equally applicable to many other music production scenarios, and could profoundly change how we produce music. In fact, AI is already having a significant impact in areas such as mastering and composition.

You Hum It, I’ll Tap It

Tamer Rashad is founder and CEO of Humtap.Tamer Rashad is CEO of Humtap, who have invested 100,000 hours of R&D with musicologists, artists, record companies and producers to develop AI algorithms for their eponymous iOS music production app. As Tamer explains, “Humtap listens to your voice and turns your hums into songs in the style of your favourite artist, in seconds and on the fly on your mobile phone.” The app records your hummed melody and chosen rhythm and applies AI algorithms to compose, arrange, perform and produce an instrumental track — all you need to do is hum your tune and select a drum track on your iPhone’s screen. You can even choose a musical genre, such as Depeche Mode or Metallica.

Many high-profile artists are collaborating with Humtap’s software developers, new genres are being continuously added and the algorithms are evolving. AI-created vocalists are some way off, although Tamer hasn’t ruled this out in the longer term, and an Android version is being developed. Tamer predicts a future whereby anyone without musical training, studio equipment, financial resources or access to music producers can produce an album — potentially within seconds, rather than a year. AI will do the production and mixing for you, all for the cost of a smartphone.

Master Class

Several companies already offer online automated AI mastering services. LANDR’s AI-powered mastering engine ‘listens’ to your unmastered song, identifies the genre and applies relevant mastering equalisation, multiband compression and other processing, all without human intervention. The processing is adaptive, responding to the needs of the song by continuously tweaking the EQ, compression and other processing tools throughout the track. Each time LANDR masters a track and listens to new music, the better it becomes, thanks to self-learning algorithms.

Pascal Pilon from LANDR claims: “We’re mastering a huge variety of music at a very professional level. LANDR has reached a point where it’s difficult for even professionals to know if a song has been mastered by a real person or by LANDR directly — we’ve reached a very high level of sophistication over the last four years. Right now we do over 330,000 songs a month, and if you think about it, that’s more songs than all studios in America combined. It’s really meeting the needs of a large community of artists.”

LANDR has established itself as a viable mastering option, especially for those who are not in a position to pay a human mastering engineer to work on their tracks.

CloudBounce also offer an AI mastering service, and recently published a white paper forecasting their future vision for AI-supported mastering, audio restoration and music production. Anssi Uimonen from CloudBounce points to the staggering volume of material uploaded to social media and video-sharing web sites: 300 hours of video are uploaded to YouTube every minute, and social media platforms such as Facebook are also hosting vast amounts of video content, typically with unprocessed and suboptimal audio.

Anssi Uimonen of CloudBounce.The potential market for audio restoration and optimisation is enormous and growing. Anssi says that his aim is to “build a robust, open network for AI audio production, enabling all content creators to make their videos, music and speech content sound the best it possibly can, automatically”. CloudBounce illustrate this idea by asking us to think of a content creator who records a video using her iPhone; the content is great, but the level is too low and there’s room reverberation. Although she doesn’t have the skills or budget to pay for a professional, she uses AI tools to quickly optimise the audio quality so the content can be published online.

It may become second nature for content creators to drag and drop audio files onto AI-powered software, or use online services to quickly remove background noise, unwanted room reverberation and audio artifacts without any user involvement. With this volume of content there simply wouldn’t be enough human audio engineers to go round.

CloudBounce began as an automated mastering service, but is evolving into a more ambitious AI-assisted music production space called dBounce.

AI Production Values

AI is thus already opening up new markets by automatically mastering, restoring and optimising audio that would otherwise not have been processed. However, the CloudBounce white paper goes significantly further, and the company are seeking to open up their current cloud-based AI processing engine into a new “community-driven ecosystem” known as dBounce.

The team’s long-term vision is to provide an online platform for AI-supported music production services. In time, the white paper envisages that users will be able to call upon AI services within dBounce to help with pretty much any music production task, including composition, mixing and mastering. As a user, you will work with dBounce as your AI ‘audio producer’, applying multiple algorithms to perform specific audio production tasks. These algorithms will progressively learn to understand different musical genres, and thus to make choices that will be sympathetic to your style. The automated EQ, compression and reverb settings that dBounce applies to your drums could thus sound very different depending on your musical genre.

A seasoned music producer could become part of this community ecosystem by teaching the dBounce algorithms specific audio production skills, such as how to EQ drums for a specific genre, for example Scandinavian folk. Other users could then rate the quality of this EQ algorithm, so the dBounce community contributes real-world feedback on the sources from which dBounce is learning, as part of the trusted ecosystem. Community members who create the dBounce AI audio producers will be rewarded through tokens when their algorithms are used, which can be traded for other dBounce AI audio production services.

Users can also build dBounce algorithms by uploading audio libraries for “feature detection” to train the machine learning models, and evaluate or rank the quality of the resulting algorithms to ensure high quality. This “call to arms” will reward users with their own trusted profile within the community and tokens to purchase other AI services. The dBounce community ecosystem could also provide a platform for plug-in manufacturers to draw upon “AI audio producers” to quickly perform production tasks such as mixing, so users could focus on other creative areas.

Dr Michael Terrell points to mastering as the “low-hanging fruit” of AI audio production, which requires little human interaction. The future road map for AI audio services is likely to become increasingly sophisticated and tailored to many different areas. Anssi adds that “Within five to seven years we are starting to see some really scary things that are very advanced in terms of production and getting the initial idea… or even getting your initial ideas on composition, based on behaviour and stuff that’s been learnt by the machine.”

The dBounce white paper also offers other examples of the ways in which ‘functional’ aspects of audio production, such as removing unwanted noise or ambience, could become automated. For example, content owners such as broadcasting companies may need to update the audio quality of their back catalogue to modern-day broadcasting standards. AI could quickly optimise this audio material, without the broadcasters needing to invest thousands of hours in repetitive manual tasks.

It seems reasonable to say many of these changes won’t happen overnight, although there’s plenty of ongoing research and development that is steadily introducing AI-supported music tools into our day-to-day lives.

A Partner Not A Replacement

LANDR’s Pascal Pilon.Without exception, all the contributors I interviewed for this feature are passionate about AI as a tool for enhancing, rather than replacing, human creativity. Many of them highlight the ways in which AI can empower people to be creative by automating the more functional aspects of music production, and by side-stepping the need to acquire difficult technical knowledge and skills. LANDR’s Pascal Pilon says: “I think it’s going to become a producer’s world. Anybody with a vision of a soundscape or specific song will be able to design it and make it a reality, without having to learn instruments or needing any dexterity in the process. It’s going to become more about visionaries of music, and I think it can only bring more quality out there.”

Tamer Rashad likewise sees Humtap as helping music fans to become active creators, rather than passive listeners, saying that presently “less than 1 percent of the population can currently contribute to music creation”.

Drew Silverstein is the founder of Amper Music.Drew Silverstein from Amper Music, meanwhile, describes their online service as an “artificial intelligence composer, performer and producer that creates unique professional music tailored to any content in a matter of seconds”. Amper’s goal is to “enable anyone around the world to be able to express themselves creatively through music — regardless of their background, access to resources or training.”

Drew explains: “Everyone has creative ideas. The challenge is not coming up with the idea, it’s figuring out how to express that idea. So if I have an idea for a painting but I’ve never been trained, as soon as I try to brush the canvas it might look terrible. It’s not that my idea was bad, it’s that my ability to express my idea was insufficient.

“If you are a non-musician we are now giving you the superpower, in essence: to be able to turn your idea into reality and to be able to make the content or the music that you’ve always envisioned, but haven’t always had the ability to do.”

Amper uses artificial intelligence to automatically generate a soundtrack to your video.

Dror Ayalon, the creator of loop-sharing environment Soundscape, sees AI as augmenting rather than replacing human capabilities. “The AI and machine learning revolution is trying to solve things that we currently don’t do — because we don’t like to do them, or it’s too costly, or because we simply can’t. I think it’s pointless to try and solve things that we already do. The future is basically the combination between human work and the machine. It’s not about replacing the human. Most of the music we love, most of the sounds we love, most of the instruments we love will generally still be in the game. It’s pointless if we don’t do it.”

Social AI

As well as revolutionising existing music-production processes, AI is likely to create new markets for audio production services, as yet unknown. Perhaps surprisingly, a large emerging market is for music fans, rather than solely musicians, producers and online content creators.

Many contributors point to a more connected future, where music consumers use AI music-creation apps to enrich their interactions with others on social media. Non-musicians could create music and collaborate with artists and friends, without any musical training, access to skilled producers, studios or money. This opens up opportunities for anyone with a passion for music, regardless of technical skills, to have a creative musical experience. Fans could propose new tracks directly to their artists, generated through their AI-supported interpretation. Tamer paints a picture of a teenage girl from a village in Africa without access to studio facilities collaborating with her favourite artist using only her smartphone. AI could widen opportunities and level the playing field for everyone.

Empowering everyone to create music and interact through social media is a major growth opportunity. AI turns passive listeners into music creators, bringing consumers much closer to artists. Several contributors draw a parallel with existing content-sharing platforms such as Instagram as the future model for connecting by creating new music, and social media is seen as offering huge potential for AI-produced music.

More daringly, Tamer Rashad suggests that AI could be used to create new tracks by artists who are no longer with us. Algorithms could analyse an artist’s entire back catalogue and identify patterns in the melodies, harmonies, instrumentation and arrangements, then generate limitless new songs inspired by that artist. Reproduction would be instrumental at least for the foreseeable future, although AI-created vocalists are seen as feasible in the longer term. Once again, all you would need is a smartphone.

What About Music Producers?

The benefits of automating repetitive, non-creative ‘functional’ activities are not only relevant to non-musicians but also to seasoned producers. As Dr Michael Terrell says, “When you start a new project there’s tons and tons of stuff you do that’s similar every time but not exactly the same. There’s all these functional aspects that, if you have AI tools to assist, could probably shave off 50 to 70 percent of the time it actually takes to mix a track. For those people being paid by the track, it’s a lot to save.”

Songwriters could likewise benefit from genre-appropriate, AI-suggested instrumentation or arrangements to help get a song over the line quickly — handy when you’ve an impending deadline. They could then use an app like Humtap to create a full demo in their desired genre and share it with the rest of the band. Musicians without the skills to mix a song could also rely upon AI-supported mixing; rather than having to learn how to use compressor controls such as ratio, threshold and release, you could simply say you want the vocals to be louder and more consistent, or that the guitar needs to be more prominent than the piano.

Tamer Rashad also suggests that AI could help artists redevelop their music in alternative genres, to reach a new audience. For example, an American pop artist could draw upon AI to regenerate an existing track in a Bollywood style.

Soundscape aims to further human creativity by pairing the musical content created by producers around the world.

AI also has an important role to play in fostering collaboration between human musicians, and this is the core aim of Dror Ayalon’s upcoming online service Soundscape. Soundscape automatically matches your musical ideas with those other musicians: you record a loop, which Soundscape analyses and synchronises to a comparable musical recording from others around the world, evaluating how closely your music style matches theirs. All you need is a single click of your mouse button and Soundscape makes the connections.

Dror Ayalon, inventor of Soundscape.You can then mould shared loops into your own music, just as others can with your musical phrases. Dror explains his motivation for Soundscape: “I was always in search of new ways to get people closer to music-making and closer together. I felt that the ability to sync people’s loops automatically is just giving me an echo that there is someone out there playing something that is as weird as maybe what I play. Maybe it’s harmonics to what I play, maybe there are other people who are doing the same thing.”

Dror summarises his experiences of connecting with others as “this fascinating moment of not feeling alone — you, your guitar, your amp at home — was something that I felt I’m really interested in. Since this is the case, your results are always surprising.”

A Little Knowledge

AI may also change the tools with which music producers do their work. As Dr Michael Terrell points out, digital audio workstation software and plug-ins are often modelled upon analogue ways of working, with a mixing console and long-established types of processing such as compression and EQ. AI raises questions around our current way of interacting with DAWs. Will the analogue metaphor make way for radically different interfaces and interaction styles? Dr Terrell points to his research, which involves automating mundane tasks such as setting level balance and basic panning and EQ in order to allow musicians to focus on being musicians. He adds: “My main motivation was to build a tool that would help people to mix music well, having seen lots of friends who really struggled to do it. One of the issues around mixing is that it’s just quite complicated. You need a good technical understanding of what the audio processors are doing, you need reasonably good versions of those tools, you need decent ears, you need a good setup at home, and it’s just very, very difficult to do.”

Dr Michael Terrell is one of the researchers behind dBounce.Dr Terrell suggests that AI tools could, in effect, act as an interface between the user and the standard DAW toolset. “An AI interface could offer simple controls that would help you mix something that actually sounded better than it would have been if you tried to do it yourself, in a lot less time. With all these tools you can build a hierarchy of controls where, right at one end, you have full automation, and as you step up, you’re effectively providing more intricate controls, but through far easier-to-understand interfaces. So, when you’re using a compressor, you have all these weird controls — you know attack, release, threshold, ratio. When you try to explain to someone what that’s doing, it’s not simple. AI could enable musicians to say what they want as an outcome, such as ‘I want to sound a bit louder’ or ‘a bit more consistent’. That way, you’re allowing people to control the software using an interface that they can easily understand.”

Drew Silverstein from Amper Music continues the theme of AI offering simpler tools. “Those [DAWs] are complicated. They take time and skill to learn, which speaks of the immense power that they hold — they’re complex for a reason. For individuals who are used to using those pieces of advanced technology, Amper’s not going to replace them. Amper is going to be a collaborative tool and a creative partner to help [people] further their existing music-making in the same way that the DAW did for them 20 years ago. Amper will do the same thing. You’ve got a whole population of individuals who look at a DAW and say it’s too much. We’re not saying ‘Let’s get rid of that knowledge and expertise as a prerequisite to be able to express your creative idea.’ If you use a DAW, great. You’ll probably be able to get more out of Amper than anyone else would, or you’ll be able to go further. But you’re not required to.”

It seems reasonable to summarise that as algorithms become more widely available, AI may enable us to define the level of control we want over our production. We can still have lots of complexity and flexibility if we want it, but we can also choose easier and more immediate results, albeit with more decisions made for us. AI could become like a human engineer who quietly sets things up and helpfully manages the technology for you, so you can focus on the music.

As Dror Ayalon says, “If you think about it, for many years, the industry emulated analogue instruments in a digital environment. The basic reason behind that was to reduce cost on one hand and to be able to do things on a massive scale. I think that people are starting to understand now that we’re not actually composing music, we’re composing data. We basically produce data pieces, but this data is artistic data that we can actually change minds and lives with — so I think you will see software starting to build tools and interfaces that will allow musicians to actually interact with that. I believe that you will be able to find tools that are heavily rooted in artificial intelligence in any studio and on the phone of any artist. I believe this age is coming real soon.”

Maximum Creativity

If AI can, hypothetically, produce an album in seconds, does this mean that musicians and producers will be out of a job, and that there is no longer a role for human creativity? That’s not the way our interviewees see it. Without exception, everyone we contacted loves human involvement in music creation. As Pascal Pilon says, “For us at LANDR, it’s essential that when we talk about the AI, we always carve the creative aspects out of the equation and leave them in the hands of the artist. We see ourselves empowering artists to make creative decisions and we focus on taking the technical elements out of the equation.”

Dror Ayalon, the creator of Soundscape, describes AI as a tool to improve musical creativity: “If these algorithms create something that is not interesting — that sounds exactly like the Beatles, or exactly like hip-hop hits from last year — they won’t succeed and get enough attention. So it’s about giving creative tools to people to make innovative music on one hand and, on the other hand, using algorithms to generate synthesized new music in the most creative way possible. No-one in the field that I know about is in it just to replicate something that was done before. Every one of us is excited about the vision, excited about creativity, and if artists and sound engineers won’t adopt what we’ve built, we failed. So [we’re] working with these people to understand how can we help them do what they do. We’re not looking to replace anyone, but build tools that can make them do amazing things. Most of the innovation that we saw in music production in the past came from innovative tools, from innovative synthesizers to innovative acoustic instruments to innovative computer music. I think there is a big opportunity here for something that is dramatically different that could change the convention about how we make, how we mix, how we produce music. I believe that I can say the music production and music composition industries are looking for such innovation.”

Drew from Amper Music adds: “Regarding the concern around diminishing creativity, I think that is an argument that some would make without looking at the whole picture, because, arguably, the fact that people now can produce something or make something without knowing how to perform and write for every instrument and perform with live players already kind of does that. We’ve just accepted the current standard as OK, but it’s the fact that a kid in his bedroom can now make a beat or a hip-hop loop or a song without knowing so many things that, 100 years ago, we would have said ‘If you don’t know this, you are not creative.’ Evolution is an unstoppable part of humanity in any field.

“I would actually argue that even today the bar to create music and produce music is quite low, and doesn’t necessarily require any skill. At one point hundreds of years ago, to write and create music you needed to know music theory and to have a pen and paper. You had to be a church musician or have a sponsor. Now, if you have a cellphone and some rudimentary app you can make music that could be a hit, just because you’ve got technology that requires very little skill or expertise.

“That being said, there is still a world in which the more skill you have and the more training you have, arguably, the stronger your ability to express your idea creatively would be. As technology advances, the technological bar necessary to express yourself drops considerably. We went from pen and paper to magnetic tape to digital to mobile — every time we do that it gets easier and easier to take your idea and turn it into reality. As a consequence, the skill set and the ability to actually do that becomes more and more democratised, and so more and more are able to join the creative class whereas they previously weren’t able to.”

Dr Michael Terrell makes the point that “One of the big reasons people make music is that they like making it. It’s not some task that you want to farm out to AI, it’s something that people actually enjoy. So what you end up doing when someone wants to do something creative like that is you build tools that help them do that more effectively. So I think this idea that it reduces creativity is actually wrong. It’s the opposite. It should free up people to actually be able to make the music they want to make, without these technical burdens that are a big limitation for a lot of people.”

AI & Careers

So how will this affect people who make their living from music production? Pascal Pilon: “When we launched this thing [LANDR] our slogan was that we work for musicians. Our view was never to prevent engineers from making a living. What we looked at was that over 95 percent of music creators could not afford the service of mastering engineers. Even those that could afford mastering engineers, would only do so now and then, for example for one album a year instead of every creation. What we believe is that every time someone creates audio, it deserves to be delivered and finished with a fighting chance, so people can look at it and say ‘Hey, there is something in there that I think I can work on and improve. It’s just ready to be shared with the people that I like.’

“I think there’s a lot mastering engineers out there who are very, very good, and who are also acting as producers in their approach. They revisit and direct the overall sound of the mix — and that, to me, is producing. And that’s great. I’m really happy master engineers act as co-producers, in that we’re just bringing productivity to artists, we’re not taking jobs. So I don’t think there’s going to be less sound engineering jobs with LANDR around. I think there’s just going to be more music. I think that’s great.”

Dr Michael Terrell talks about future opportunities for professional producers to train dBounce to become an AI audio producer. “One of the things we’ve tried to do with the design of this ecosystem is that if you’re a music producer and you’re very good, you will still be — you can actually take part in this ecosystem. If you can train AI and you own it, you can earn money from that AI, so it becomes an ancillary revenue stream. If you can dump all your AI knowledge into a machine and then earn the money from that, you can probably enjoy your life doing whatever you want to do. So it’s about getting involved with the ecosystem.”

Drew Silverstein from Amper Music again draws an analogy with earlier technologies. “When the synthesizer came out, everyone worried that there would never be another live musician ever again. Orchestras are still around, and some are thriving more than ever. Depending on what area you put yourself in, there’s always going to be this existential concern about the future. What we’ve seen is that for those who take advantage of the opportunities provided to them by advancing technology, the future in some ways is never brighter.”

Drew draws a distinction between artistic and functional music. “We look at media music in two different buckets. One we call artistic music and one we call functional music. Artistic music is valued for the collaboration and creativity that goes into making it. It’s about the creative process of working with another person to make art that’s artistic. Our perspective is that that will never go away, even when Amper and AI music is indistinguishable from human-created music. This is because the value is not about the final product, it’s about the process of making it. It’s about [John] Williams and [Steven] Spielberg sitting down and scoring the next Star Wars. It’s that artistic mind-melding of multiple individuals making art.”

Functional music, by contrast, is “music that exists in a world where it serves a purpose — but no-one gets in an elevator and asks about the music, what it’s doing and why it’s there. Functional music is valued for the purpose that it meets, but not for the creative process that went into making it. In those situations individuals who [need] functional music often have creative ideas themselves — musical ideas. We aim to help them directly turn their idea into reality.

“And so what I say to people who are concerned about jobs is a couple of things. One, your job in five years will probably not exist. Your career certainly will, and it’s important those two things don’t get separated out of context — because otherwise it can make me look terrible when I don’t intend it to be! What I mean is the way that we make a living in music has changed for all time as technology advances. As long as we’re able to adapt to the advances and to take advantage of what technology offers us, then we’ll be able to continue providing value to others. We’ll be able to keep our career, but that doesn’t require that our job remains the exact same, because it’s not and it shouldn’t be. If it was then where would progress be? We would be stuck in whatever era you want to define as the era we should be stuck in. So part of the negative reaction to this evolution is fear based on an uncertainty within which an individual exists: ‘I’ve spend my entire life trying to do this thing. What happens to me in this future that as of now is very opaque, and because of that, is very concerning to me.’”

Drew adds that “while Amper will change the future of music, it won’t replace it. As long as an individual is creating music which is valued artistically that is delivering value to whoever it is — whether that is a listener or a client or creator — in the way that they need value to be delivered, then humans will be able to create and make a living from music or other art for all time.”

Similarly, it seems plausible that the best producers will always be in demand. It looks likely that AI will not replicate all the interactions and deep working relationships between artist and producer any time soon. In essence, people like people.

Too Much Music?

If AI apps empower everybody to generate music, is there a risk that great music will simply become lost in huge amounts of online content? Pascal Pilon compares music with video content: “If you think about the quality of lenses today and what people are using from their phones, there’s a ton of pictures being taken and there’s a lot of videos being made. Are we saying that there are too many videos? Not so much, we have the attention span that we do. YouTube is full of videos, and we watch what we want to watch. I think music is becoming something we access via recommendations from friends and people. I think the way we navigate the offering will change.”

Pascal also points out that AI tools can help navigate the flood of music, helping us to find artists to listen to or work with. “In terms of sounds that you’re looking for, in terms of collaborators you want to work with, in terms of listeners you should be listening to, I think AI will make that process very efficient. Making music and getting heard will become more like posting on Instagram and getting followers. Everything’s going to be residing more on social networks, and I think the music landscape will trend towards the image landscape — very fragmented, very wide. You’re still going to be able to buy professional photography at $20,000. That hyper quality will still be featured in a certain way, but listeners’ tastes will prevail. I don’t think that it matters if there’s five million songs or five billion songs out there. I mean most of those songs will be only important to the artists themselves and their immediate relatives. The best will stand out, and the methods by which we find them will be very exciting. Our relationship to music will be more social-media based, as will our response to music.”

It’s Up To Us

The future is likely to see an explosion of new music created by people who previously did not have the means to do so, enabled by AI-based tools. Equally, musicians and producers are likely to find more time to be creative and to create music in new ways, leaving AI to manage repetitive functional tasks. Creating music is likely to become part of people’s online interactions, with consumers creating and sharing new music through social media platforms.

AI for music production and consumption is a huge area, and this article can only offer a flavour of ongoing research and development for AI music services. As Cliff Fluet puts it, “AI is going to become a term that’s as helpful as talking about ‘the Internet’. Twenty years ago, if I’d tried to describe the Internet, you would think it was just one platform — and actually it was the first step to something huge.”

The rich opportunities around AI raise many questions, and to date, we only have some answers. Many developments are in their infancy and some timings remain uncertain. What is certain is that AI is receiving serious attention from major record labels, number one artists, business incubators and leading-edge software developers, so changes are on their way. Many contributors estimate we’ll start to see a shift in music production and consumption within the next five to seven years. Widespread adoption will come down to how easily we can interact with our AI tools and the quality of the end results, which are steadily advancing.

As with any emerging technology, we clearly have a choice about whether to use it and how we want to use it, and it may be the market that shapes the direction of AI software. All of our contributors were keen to emphasise that AI will not replace artists or producers. Paradoxically, the more we use AI within music, the more we may draw people together.

The Future Of The Music Business

If AI changes the music-industry landscape, how might industry bodies such as major record labels respond?

Cliff Fluet of media advisory firm Eleven.Cliff Fluet, Managing Director from media technology advisory firm Eleven, says: “Because there are so many different types of AI music tools, the response from major labels really depends on what the technology is and what it does. To shape-shift [change the genre of] tracks or make them adapt or remix them at a vast speed is extremely interesting to them. These are tools that make what was a very manual, costly and laborious process fast, cheap and efficient. In a world where everything’s run by streaming web site playlists, they can potentially create a version of a song for every genre — to have a version in every playlist, that’s money for them. For sync, labels make a lot of money from using music in ads and for brands, so AI technology that adapts existing recordings more easily is also popular.

“There is much excitement, but a certain hesitation to ask incumbent labels to embrace pure generative AI. There’s Upton Sinclair’s quote — “It is difficult to get a man to understand something if his salary depends on not understanding it” — and that very much is applicable to generative AI music. If you live in a world where musicians and/or A&R people felt that a machine could just create a performance… one might say they’re intellectually invested in not taking it up.

“Taking the long view, because I’m really, really old… I’ve seen all this before. There was word in the ’70s that synthesizers meant there would be no more pianists or keyboards, or drum machines meant no more drummers. There was then a word that music sampling would destroy music, or that putting digital music online would mean the end. Not only did none of those things turn out to be true, but the exact opposite turned out to be true.”

LANDR’s Pascal Pilon points out that “The music industry is a laggard in many aspects. Look at the financial market and how start-ups around the world are seeking financing to develop and commercialise products with the assistance of venture capital partners. In the music industry, this role has been played historically by the record label, but sourcing a deal with labels is still very rudimentary. I think the music industry’s financing model will progressively become more data-driven, more competitive. It will draw more capital from a wider set of investors, still involving labels but also a larger set of small to large angel investors.”

Pascal predicts that music labels will still need to invest heavily in talent to promote music. “Even if there’s AI, you will still need people to bring artists visibility. Think marketing: the rarest thing you get is listeners in the music space that spend. Your ability to bring people to listen to you will still require a cast of collaborators from a marketing perspective. Again, it’s very similar to the movie industry. Look at how important the production budget is — yes they’re huge, but think about the commercial budget, which is even bigger. To gain people’s attention, you will need to spend on channels to deliver that — think Facebook. If you can bring virality to the likes of Google and YouTube then you will be able to be independent, but it will require lots of sophistication from a promotion standpoint.”

AI & Copyright

Who owns music that is generated by artificial intelligence? And if AI software can be inspired by existing songs and generate new music of a similar style, does copyright become a grey area?

Like every new area of technology, AI promises to generate plenty of work for lawyers.Cliff Fluet of media advisory firm Eleven insists: “It’s not grey at all. It’s that people don’t like to hear the answer, depending on what they’re doing. But you’ve asked the key question and there’s a whole bunch of lawyers speculating on this stuff!

“This is as good or as bad a question as asking who owns a movie. The real answer is that it kind of depends on how you made it, how much of it is yours and how much of the stuff is owned by somebody else. To what extent are you getting the input of actors, set designers, third parties and so on? To what extent did you create something wholly original? The answer is as simple and sophisticated as that really. Every single one is dependent on how their processes, how their algorithms, how their choices are made. If you’re a really good lawyer the answer is very, very simple, provided you really understand the depths of the processes.”

Given that AI for music production is an emerging technology, is this a legally tested area?

Cliff Fluet: “I’d say it is and it isn’t. The law on music copyright has been built upon highly unreliable technology called musicians and the human brain. You’ve got people asking ‘What if a machine replicates something else?’ — and it’s like ‘Well, what if an artist replicates something else?’, which is what happens every single day in the music industry. There’s real questions about did someone actually hear something or did they not? Did someone actually adapt something? The law is not different when it’s a machine or human being. The difference between a human being and a machine is that with a machine you have absolute evidence.”

Cliff points to previous legal cases where “the jury has to sit there and try and work out whether someone heard something, adapted something or did it. With a machine, you don’t have to rely on its memory, you just literally can see all of the notes there, so actually I think arguably it’s easier rather than more difficult. Working out whether or not a human has ripped off a composition or whether or not they own it, or whether or not they shared it, is far, far more complex than a machine. If anything, the advancements in technology could help.” source

Robots can’t be creative, right? A number of startups that focus on the development of artificial intelligence that is capable of composing music have proved this statement wrong. AI music composers generate original, copyright-free music you can use in your latest YouTube video or social media video ad. You don’t even have to be a sound designer or a musician to produce soundtracks for your videos with AI Music Composers, because you can upload the music that has already been recorded and create variations of it. So, let’s take a look at what AI music composers have to offer in 2023.

AI MUSIC GENERATORS AND TOOLS TO KNOW

Voice.ai

Topmediai.com

Play.ht

Azure.microsoft.com 

Murf.ai

Vall-e.io

Bottalk.io

uberduck.ai

Voicemod.net

Veritonevoice.com

Podcastle.ai

Resemble.ai

My Own Voice

Lovo.ai

Descript.com

Cereproc.com

Whisper

Spik.ai

Respeecher

Speechify

Speechelo

Synthesys.io

Spik.ai

Bigspeak.ai

Replica

Woord

Clipchamp

Voicera

Natural Reader

Search our site for any articles about voice cloning here

below check out our other articles over the years on the topic:

Microsoft’s New AI Can Simulate Anyone’s Voice From a 3-Second Sample

The Rise Of Voice Cloning And DeepFakes In The Disinformation Wars

What is Voice Cloning and How Does It Work?

Ai Voice Cloning : How does it work and where is it used?

Copy That: Realistic Voice Cloning with Artificial Intelligence

The Era of Voice Cloning: What It Is & How to Get Your Voice Cloned

What is Voice Cloning and How Does It Work?

So you keep seeing Midjourney all of a sudden, but what exactly is it? Let’s take a quick look at what Midjourney is, how to use Midjourney, and how much Midjourney costs.

The service can create a variety of art styles, from realistic to abstract, and is highly regarded for its detailed and highly rendered images.

What is AI Art Generator ?

An AI art generator is a software tool that uses artificial intelligence to create original visual art. It uses neural networks trained on large datasets of images and art. By analyzing text prompts from users, the Muse AI art generator produces new images matching the described artistic style, composition, and qualities. This allows anyone to easily create unique digital artwork just by typing what they want the AI to generate.

How Does Midjourney Work?

Midjourney, as of now, is closed-sourced and self-funded, so the exact mechanics are not public. However, like other popular AI tools, Midjourney uses different machine-learning technologies, including large language and diffusion models.

Like other AI tools, due to the way it collects and uses data to create new images, there are ethical debates and allegations of both art theft and copyright infringement.

About

Midjourney is an independent research lab exploring new mediums of thought and expanding the imaginative powers of the human species.

We are a small self-funded team focused on design, human infrastructure, and AI. We have 11 full-time staff and an incredible set of advisors.

Midjourney is a generative artificial intelligence program and service created and hosted by San Francisco-based independent research lab Midjourney, Inc. Midjourney generates images from natural language descriptions, called “prompts”, similar to OpenAI’s DALL-E and Stability AI’s Stable Diffusion.

Midjourney Topics:

Text To Image with AI Art Generator

Play with the AI below https://www.midjourneyai.ai/

Join their Discord Server below https://discord.com/invite/midjourney

What is voice cloning?

Voice cloning is the name of the process that allows artificial intelligence to create a cloned or copied version of a natural human voice.

When done well, the natural and synthetic voices are confusingly similar. Often this cloning model is applied to voices for text-to-speech technologies.

More commonly known is the term voice cloning as “Deepfake voice”.

How exactly does cloning a voice with AI work?

When a speaker wants to clone his voice synthetically, he must first record a large amount of audio material. This is used to create a dataset of all the nuances of his voice. Certain audio specialists develop the cloned voice through a network of virtual neuron connections. The process takes a lot of time and effort – yet the results are baffing.

Step 1 – speaker selection

The central component is an extensive collection of audio samples of human speech, and it is important that these audio recordings are of high quality. We recommend choosing a speaker who has experience with this type of recording and having the recordings done by a sound engineer with professional equipment.

Have in mind that the target group likes the selected voice and speaking style – it will represent your company.

Step 2 – transcript creation

As said the cloning model often is applied to voices for text-to-speech technologie. In this case the transcript is the written part and necessary for the voice cloning. A transcript is a collection of sentences and utterances that must be spoken by the speaker.

If you produce a voice clone with BotTalk, we will take care of the transcript based on your content.

Step 3 – recordings

Each audio file should contain a single sentence, which is delivered in the transcript. All files must be in the same spoken language and style. We recommend defining a speaker persona beforehand. This document defines elements such as the characteristics of the voice and the character behind the voice. This helps to guide the process of creating a voice clone and your expected result. Important is that the recordings must match the corresponding transcript by 100%. Errors in the transcripts will lead to loss of quality during the training.

Step 4 – processing

When the data set of the transcript and recordings are ready, the processing can begin. Audio specialists will process through a neural network and create a voice clone. After every iteration, you will receive a voice sample to review the quality of the voice. If the quality is too low, you provide fewer voice data, and the audio specialist team will need further recordings. Through this process, they can directly participate in the development and follow the training of the voice clone.

Where is voice cloning used today?

Film & Television Industry

The voices of actors can be synthesized to facilitate the work of productions. Primarily when an actor records audio for marketing and advertising purposes, their voice clone can be used much more widely. This process saves time and costs and allows the possibility to use that voice in 10 years.

In the last Star Wars movie, the film editors had to painstakingly recreate the character of General Leia Organa after actress Carrie Fisher died during filming.

Imagine how much easier this would be with voice cloning.

Publishiers & News

At BotTalk we produce voice clones for radio stations and publishers.

Our goal is, on the one hand, to develop new voices for accents and dialects that do not exist on the market. On the other hand, to make it possible for publishers or authors to establish a unique brand AI-voice to promote digital content further.

With voice cloning the editorial team can read aloud every single article in your local newspaper. The synthetic voice will sound much more familiar to the listener than a standard synthetic voice. A similar use case is applicable for radio stations to read aloud the weather forecast or traffic information by a synthetic voice.

Education

Voice clones are often used to make life easier for online educators..Their voices can be synthesized to facilitate the work of productions. Once enough audio material is available, the professor can further dub his videos with the help of TTS (and the voice clone) with little effort.

source

Create Your Own AI Animated Avatar: A Step-by-Step Guide

• Links: ChatGPT for script creation: https://chat.openai.com/chat
• MidJourney for image generation: https://www.midjourney.com
• Elevenlabs for audio generation: https://beta.elevenlabs.io/
• D-ID for video generation: https://studio.d-id.com/

Have you ever wished you could clone your voice? Now there’s no need to wish, because the era of voice cloning has arrived! In this blog post, we’ll take a look at what voice cloning is and how you can get your very own voice clone. So read on to find out more!

What Is Voice Cloning?

Voice cloning is the process of creating a digital copy of someone’s voice. In the past, this could be done using voice recognition software or by recording the person’s voice and then using that recording to create a digital copy. Today, however, it is possible to get your voice cloned really easily with an AI-powered platform like Podcastle. But before we tell you how Podcastle works, let’s understand how long has voice cloning been around?

The truth is voice cloning has been around for many years. The first voice cloning technology was created in 1998, and since then the technology has continued to develop and improve.

In general, if you have really strong technical skills there are a number of ways to create a voice clone. One of the most popular methods is to use voice recognition software. This software can be used to create a digital copy of someone’s voice by recording them speaking and then analyzing their speech patterns.

Another way to create a voice clone is by recording the person’s voice and then using that recording to create a digital copy. This method involves manually creating a digital copy of the person’s voice by cutting out snippets of their speech and then piecing them together like a puzzle.

Both of these methods require some technical skill, but there are also a number of services that offer cloned voices for hire. These services use professional voice actors to recreate someone’s voice, so they can be used for audio books, phone greetings, or any other purpose where an exact replica of someone’s voice is required.

The best news is that now you don’t even need to spend a fortune on hiring professional voice actors to recreate your voice or the voice of someone else, which honestly, will not even sound too similar to the original voice. In this article, we’ll tell you how to clone your voice yourself using Podcastle’s voice cloning technology.

But why would you really want to get your voice cloned? What are the benefits of voice cloning? Let’s find out:

Podcastle Team

In the second installment to our deepfake voice series, we will explore voice cloning by defining what it is, how it works, and the benefits of this new technology. As we progress through this series, we’ll delve into more detail on synthetic voice content, how you can create one, and ways to address the issues of ethical use and deepfake fraud.

In this blog, we’ll be covering these following areas:

• What is voice cloning?
• How is voice cloning possible?
• What are the benefits of voice cloning?
• What is the best voice cloning application?

What is voice cloning?

Voice cloning is often tossed around with other terms, such as deepfake voice, speech synthesis, and synthetic voice, that have slightly differentiated meanings. Voice cloning is the process in which one uses a computer to generate the speech of a real individual, creating a clone of their specific, unique voice using artificial intelligence (AI).

Text-to-speech (TTS) systems, which can take written language and transform it into spoken communication, is not to be confused with voice cloning. TTS systems are much more limited from the outputs they produce compared to voice cloning technology, which is really more of a custom process.

With a TTS system, the training data, the key component to any synthetically created media, informs the production of a voice output. In other words, the voice you hear is the one that was given in the data set.

Now, with the introduction of voice cloning AI technology, that changes. Methods have been put in place to provide deeper analysis and extraction of the characteristics of a target voice. These attributes can then be applied to different waveforms of speech, allowing someone to change the speech output of one voice to another.

How is voice cloning possible?

Thanks to advancements in artificial intelligence (AI), particularly deep learning, a subset of machine learning underneath the umbrella of AI, we’ve been able to produce accurate replications of a voices. But this only made possible by two things:

• Powerful hardware with cloud computing capabilities to process and render in a timely and efficient manner
• Extensive training data of the targeted voice from which models can leverage to create an accurate voice clone

With the proper AI and developmental expertise and tools, it really comes down to the latter. You need a large amount of recorded speech to have enough data to train the voice model. The information around the voice is stored in an embedding, a fairly low-dimensional space where you can translate discrete variables into high-dimensional vectors.

In other words, it makes it easier to work with large inputs with machine learning models. For the sake of not getting too technical, we’ll leave it at that, but feel free to dive deeper into the subject if that interests you.

What are the benefits of cloning your voice?

Let’s start with the good. There are plenty of potential use cases for voice cloning that often become overshadowed by the negative uses, which we will address in a second. Some of the positive applications of technology include:

• Increase advertising and sponsorship opportunities for voice personalities, celebrities, and influencers
• Help companies work with talent during their busiest times of the year, such as football season for players or coaches
• Revive voices from the past for use in entertainment to help tell a story in documentaries, movies, and TV shows
• Diversify broadcast content for repeat content such as weather reports or sports updates
• Localize content so that it can be heard in the host or narrators voice in another language

These are just some of the positive uses for voice cloning, and as the technology continues to evolve, more will emerge. But of course, everything hinges on the ethical use of someone’s voice. That’s why the need for a movement towards the standardization of the approval process is so imperative to protect everyone’s voice and ensure they have complete control over how it’s used.

What’s the best voice cloning app?

To narrow down your search to find the best voice cloning applications you should first determine what you are looking for. Do you need something that’s more for text-to-speech output? Or do you need something more custom?

Once you’ve figured out why you need a voice cloning application, you should then hone in on three key criteria:

1. Output quality: you’ll want to make sure that the output is authentic sounding and meets your prescribed needs. Usually, they will have samples of what the product can do. If not, you should consider asking for a demo, if available, to determine how human their product sounds.
2. Intuitive interface: how easy is it to use the application? Is it hard to find things when you’re in the app or can you navigate and use it to meet your needs? Again, this can be determined by product videos, marketing content, and a demo.
3. Voice protections: you’ll want to make sure that the company follows ethical uses of voices. If it’s a custom service requiring training data, then it’s important to inquire about data protections and how a voice, when created, won’t be used improperly.

The ethical implications around voice cloning are the nexus of Veritone Voice, our voice-as-a-service application. Built within the framework of the application are the levers to give users control over their voice, enabling the proper protections so that they decide who can use their voice. This helps us deliver our custom voice-as-a-service solution to enable a complete white glove experience for the talent we work with.

In the next chapter in this series, we’ll be discussing Text-to-Speech (TTS) AI and how it’s related to voice cloning.

ETHAN BAKERDIRECTOR OF CONTENT DEVELOPMENT, VERITONE

What is Voice Cloning and How Does It Work?

Voice cloning is a technology that imitates a person’s voice and replicates it to use for voice assistance. Traditionally, cloning a voice requires hours of recorded speech to build a collection of datasets that can be used to build a new voice model. However, now it’s possible in seconds!

With Voice.ai, it’s possible in a matter of seconds. Voice.ai’s Voice Universe users have captured high quality voices to build a library of over 4000 user-generated characters. As a result, the software can analyze, modulate and correct anyone’s voice before turning it into a pre-selected, A-list celebrity impression in real-time. Voice cloning can enhance your live-stream, group chat or gaming experiences like never before.

An Ever-Growing Library of Voices!

Try the Voice AI voice changer and get ready to be amazed by how realistic the voices sound. We use deep learning technology to create realistic AI voice clones that sound exactly like an impression of anyone in the world; whether it’s your favourite reality star, game character, celebrity or cartoon character. Our Voice Universe community of contributors are training incredible voices every day to create hyper-realistic replicas that sound exactly like the real person! Browse through our ever-expanding library and flick through different voices to sound scarier, funnier, younger, older, smarter or just plain obnoxious – all by simply changing the AI voice character.

The Voice.ai real-time AI voice changers works for all sorts of games: Among Us, World of Warcraft, Minecraft, CS:GO, League of Legends, PUBG, Rust, GTA V, Second Life, Valorant and applications: Discord, Skype, Google Meet, Zoom, WhatsApp, Teamspeak, OBS and practically every other Windows app or client!

Voice.ai

The development of a medical triage program raises a question: When lives are at stake, should artificial intelligence be involved?

[The U.S. says humans will always be in control of AI weapons.
But the age of autonomous war is already here.]

source

By Pranshu Verma

AI could replace human military commanders in making life or death decisions on the battlefield, DARPA suggests

• DARPA is working with decision makers to train algorithms to make decisions
• The idea is that humans have a bias and can disagree, slowing down decisions
• AI can be trained from the start, based on best practice, to make fast decisions
• The technology is still at the early stages, but DARPA hopes for a wide roll out 

Modern military operations, whether it be combat, medical or disaster relief, require complex decisions to be made very quickly, and AI could be used to make them.

The Defense Advanced Research Projects Agency (DARPA) launched a new program aimed at introducing artificial intelligence into the decision making process.

This is because, in a real world emergency situation, that might require instant choices between who does and doesn’t get help, the answer isn’t always clear and people disagree over the correct course of action – AI will make a quick decision.

The latest DARPA initiative, called ‘In the Moment’, will involve new technology that could take difficult decisions in stressful situations, using live analysis of data, such as the condition of patients in a mass-casualty event and drug availability.

It comes as the U.S. military increasingly leans on technology to reduce human error, with DARPA arguing removing human bias from decision making will ‘save lives’.

The new AI will take two years to train, then another 18 months to prepare, before it is likely to be used in a real world scenario, according to DARPA.

‘AI is great at counting things,’ Sally A. Applin, an expert in the interaction of AI and ethics, told Washington Post, adding ‘I think it could set a precedent by which the decision for someone’s life is put in the hands of a machine.’

According to DARPA, the technology is only part of the problem when it comes to switching to AI decision making, the rest is on building human trust.

‘As AI systems become more advanced in teaming with humans, building appropriate human trust in the AI’s abilities to make sound decisions is vital,’ a spokesperson for the military research organization explained.

By Ryan Morrison source

When Was Artificial Intelligence Invented?

Advances in artificial intelligence (AI) have given the world computers that can beat people at chess and “Jeopardy!,” as well as drive cars and manage calendars. But despite the progress, engineers are still years away from developing machines that are self-aware. Some believe the resulting technological singularity will eradicate poverty and disease, while others warn it could endanger human survival.

Can Machines Think?

In the first half of the 20^th century, science fiction familiarized the world with the concept of artificially intelligent robots. It began with the “heartless” Tin man from the Wizard of Oz and continued with the humanoid robot that impersonated Maria in Metropolis. By the 1950s, we had a generation of scientists, mathematicians, and philosophers with the concept of artificial intelligence (or AI) culturally assimilated in their minds. One such person was Alan Turing, a young British polymath who explored the mathematical possibility of artificial intelligence. Turing suggested that humans use available information as well as reason in order to solve problems and make decisions, so why can’t machines do the same thing? This was the logical framework of his 1950 paper, Computing Machinery and Intelligence in which he discussed how to build intelligent machines and how to test their intelligence.

Making the Pursuit Possible

Unfortunately, talk is cheap. What stopped Turing from getting to work right then and there? First, computers needed to fundamentally change. Before 1949 computers lacked a key prerequisite for intelligence: they couldn’t store commands, only execute them. In other words, computers could be told what to do but couldn’t remember what they did. Second, computing was extremely expensive. In the early 1950s, the cost of leasing a computer ran up to $200,000 a month. Only prestigious universities and big technology companies could afford to dillydally in these uncharted waters. A proof of concept as well as advocacy from high profile people were needed to persuade funding sources that machine intelligence was worth pursuing.

The Conference that Started it All

Five years later, the proof of concept was initialized through Allen Newell, Cliff Shaw, and Herbert Simon’s, Logic Theorist. The Logic Theorist was a program designed to mimic the problem solving skills of a human and was funded by Research and Development (RAND) Corporation. It’s considered by many to be the first artificial intelligence program and was presented at the Dartmouth Summer Research Project on Artificial Intelligence (DSRPAI) hosted by John McCarthy and Marvin Minsky in 1956. In this historic conference, McCarthy, imagining a great collaborative effort, brought together top researchers from various fields for an open ended discussion on artificial intelligence, the term which he coined at the very event. Sadly, the conference fell short of McCarthy’s expectations; people came and went as they pleased, and there was failure to agree on standard methods for the field. Despite this, everyone whole-heartedly aligned with the sentiment that AI was achievable. The significance of this event cannot be undermined as it catalyzed the next twenty years of AI research.

Roller Coaster of Success and Setbacks

From 1957 to 1974, AI flourished. Computers could store more information and became faster, cheaper, and more accessible. Machine learning algorithms also improved and people got better at knowing which algorithm to apply to their problem. Early demonstrations such as Newell and Simon’s General Problem Solver and Joseph Weizenbaum’s ELIZA showed promise toward the goals of problem solving and the interpretation of spoken language respectively. These successes, as well as the advocacy of leading researchers (namely the attendees of the DSRPAI) convinced government agencies such as the Defense Advanced Research Projects Agency (DARPA) to fund AI research at several institutions. The government was particularly interested in a machine that could transcribe and translate spoken language as well as high throughput data processing. Optimism was high and expectations were even higher. In 1970 Marvin Minsky told Life Magazine, “from three to eight years we will have a machine with the general intelligence of an average human being.” However, while the basic proof of principle was there, there was still a long way to go before the end goals of natural language processing, abstract thinking, and self-recognition could be achieved.

Breaching the initial fog of AI revealed a mountain of obstacles. The biggest was the lack of computational power to do anything substantial: computers simply couldn’t store enough information or process it fast enough. In order to communicate, for example, one needs to know the meanings of many words and understand them in many combinations. Hans Moravec, a doctoral student of McCarthy at the time, stated that “computers were still millions of times too weak to exhibit intelligence.”  As patience dwindled so did the funding, and research came to a slow roll for ten years.

In the 1980’s, AI was reignited by two sources: an expansion of the algorithmic toolkit, and a boost of funds. John Hopfield and David Rumelhart popularized “deep learning” techniques which allowed computers to learn using experience. On the other hand Edward Feigenbaum introduced expert systems which mimicked the decision making process of a human expert. The program would ask an expert in a field how to respond in a given situation, and once this was learned for virtually every situation, non-experts could receive advice from that program. Expert systems were widely used in industries. The Japanese government heavily funded expert systems and other AI related endeavors as part of their Fifth Generation Computer Project (FGCP). From 1982-1990, they invested $400 million dollars with the goals of revolutionizing computer processing, implementing logic programming, and improving artificial intelligence. Unfortunately, most of the ambitious goals were not met. However, it could be argued that the indirect effects of the FGCP inspired a talented young generation of engineers and scientists. Regardless, funding of the FGCP ceased, and AI fell out of the limelight.

Ironically, in the absence of government funding and public hype, AI thrived. During the 1990s and 2000s, many of the landmark goals of artificial intelligence had been achieved. In 1997, reigning world chess champion and grand master Gary Kasparov was defeated by IBM’s Deep Blue, a chess playing computer program. This highly publicized match was the first time a reigning world chess champion loss to a computer and served as a huge step towards an artificially intelligent decision making program. In the same year, speech recognition software, developed by Dragon Systems, was implemented on Windows. This was another great step forward but in the direction of the spoken language interpretation endeavor. It seemed that there wasn’t a problem machines couldn’t handle. Even human emotion was fair game as evidenced by Kismet, a robot developed by Cynthia Breazeal that could recognize and display emotions.

Time Heals all Wounds

We haven’t gotten any smarter about how we are coding artificial intelligence, so what changed? It turns out, the fundamental limit of computer storage that was holding us back 30 years ago was no longer a problem. Moore’s Law, which estimates that the memory and speed of computers doubles every year, had finally caught up and in many cases, surpassed our needs. This is precisely how Deep Blue was able to defeat Gary Kasparov in 1997, and how Google’s Alpha Go was able to defeat Chinese Go champion, Ke Jie, only a few months ago. It offers a bit of an explanation to the roller coaster of AI research; we saturate the capabilities of AI to the level of our current computational power (computer storage and processing speed), and then wait for Moore’s Law to catch up again.

Artificial Intelligence is Everywhere

We now live in the age of “big data,” an age in which we have the capacity to collect huge sums of information too cumbersome for a person to process. The application of artificial intelligence in this regard has already been quite fruitful in several industries such as technology, banking, marketing, and entertainment. We’ve seen that even if algorithms don’t improve much, big data and massive computing simply allow artificial intelligence to learn through brute force. There may be evidence that Moore’s law is slowing down a tad, but the increase in data certainly hasn’t lost any momentum. Breakthroughs in computer science, mathematics, or neuroscience all serve as potential outs through the ceiling of Moore’s Law.

The Future

So what is in store for the future? In the immediate future, AI language is looking like the next big thing. In fact, it’s already underway. I can’t remember the last time I called a company and directly spoke with a human. These days, machines are even calling me! One could imagine interacting with an expert system in a fluid conversation, or having a conversation in two different languages being translated in real time. We can also expect to see driverless cars on the road in the next twenty years (and that is conservative). In the long term, the goal is general intelligence, that is a machine that surpasses human cognitive abilities in all tasks. This is along the lines of the sentient robot we are used to seeing in movies. To me, it seems inconceivable that this would be accomplished in the next 50 years. Even if the capability is there, the ethical questions would serve as a strong barrier against fruition. When that time comes (but better even before the time comes), we will need to have a serious conversation about machine policy and ethics (ironically both fundamentally human subjects), but for now, we’ll allow AI to steadily improve and run amok in society.

Rockwell Anyoha is a graduate student in the department of molecular biology with a background in physics and genetics. His current project employs the use of machine learning to model animal behavior. In his free time, Rockwell enjoys playing soccer and debating mundane topics. this story had some contributions by Rockwell Anyoha

Click Here for a Complete Historical Overview
from University of Washington in PDF format
that is different from the material seen on this page