machines cannot be as creative as humans

December 01, 2020 | mins read

Gil Weinberg investigates this creative … In research presented at NIPS last year, researchers from Microsoft generalized the above interpretation of GANs to a broader class of games, providing a deeper theoretical understanding of the GAN learning objective and enabling GANs to apply to other machine learning problems such as probabilistic inference. Both the generator and the adversary play the same game but with different goals: the generator tries to fool the adversary, and the adversary tries to remain accurate in identifying samples from the generator. The generator network transforms a set of random numbers through a neural network into an observation, such as an image. Humans can think in a way that no computer will ever be able to match let alone imitate convincingly. He looks into how one may in future be able to program consciousness into a computer akin to the way that we evolve consciousness. 2. creativity is accomplished by problem solving. As a result, GANs summarize the distribution via a low-dimensional manifold but do not accurately capture the full training distribution. But what comes next? However, one key difficulty in training GANs is the issue of dimensional misspecification: a low-dimensional input is mapped continuously to a high-dimensional space. They will acquire more knowledge by scouring the internet than we can gain in a lifetime, and by experiencing emotions vicariously will be able to convince themselves and us that they have actually had these human experiences. As readers likely remember, an artificial intelligence known as AlphaGo defeated Lee in 2016. Machine learning models trained on images need to capture this multileveled structure to effectively reason about image content. Even though we're no longer consciously thinking about the problem, the passionate desire to solve it keeps it alive in the unconscious where it can be mulled over freely and uninhibited in ways not always possible with conscious thought. All this can lead hopefully to an illumination, as the solution to the problem bubbles up into consciousness. In 1908 the French mathematician, philosopher and scientist Henri Poincaré suggested a four-stage cycle: conscious thinking, unconscious thinking, illumination and verification. He argues that, in general, machines can only do what they’ve been programmed to do—we can program a machine to create a picture, and in that sense it is creative; but we still have to tell it how to create by giving it … Now one would imagine that as machines lack imagination, they are not creative. What powers the creative urge? Recent methods in artificial intelligence enable AI software to produce rich and creative digital artifacts such as text and images painted from scratch. The machine went beyond its training set to create images that no one had ever dreamt of before. The technique’s result shows that by making training data more variable through adding noise, one really encodes a preference that the learned machine learning system should depend in a smooth way on its input. GANs learn to generate content by playing a two-player game between a generator network and an adversary network. A particularly advanced set of machines could replace humans at literally all jobs. SALON ® is registered in the U.S. Patent and Trademark Office as a trademark of Salon.com, LLC. They can then seek complex correlations … One group of researchers had tried tinkering with the connections between the machine's mathematically-simulated neurons so as to generate something that resembled a cat at each layer of neurons. Certainly AlphaGo showed creativity. One thing is certain: their creativity will be unlimited. Through unconscious thought, he had found a way to use computer vision to explore how artificial neural networks worked. In the not too distant future machines will have the ability to read a language fluently. When a human being makes a leap forward and produces something that goes beyond the initial material, we call it creativity. ------------------------------------------, The Artist in the Machine: The World of AI-Powered Creativity. In this case both human and machine showed creativity. How does the machine see the world? In the work presented today at NIPS 2017, the problem of noise variance is overcome, as presented in a paper entitled Stabilizing Training of Generative Adversarial Networks through Regularization by Kevin Roth, Aurelien Lucchi, Sebastian Nowozin and Thomas Hofmann. Machines actually have unpredictability built in. The technical contribution of the work is to derive an analytic approximation to the addition of noise and showing that this corresponds to a particular form of regularization of the variation of functions. In the following figure, we show a comparison of a network generating face images using the ResNet architecture. But could machines achieve such breakthroughs? Machines are redefining what it is to be living, not merely … Initially proposed by Ian Goodfellow and colleagues at the University of Montreal at NIPS 2014, the GAN approach enables the specification and training of rich probabilistic deep learning models using standard deep learning technology. Reproduction of material from any Salon pages without written permission is strictly prohibited. The grandmaster later commented that AlphaGo had displayed "human intuition.". Toward this end, Microsoft Research is committed to the research to make GAN models more practical and applicable to more areas of study. Lectures from Microsoft researchers with live Q&A and on-demand viewing. Awareness: The key difference between human and machine creativity. One must fight for recognition of one's ideas. In a sense, it was not possible for machine learning models to be creative and create complex observations such as entire images. The great question: What makes us creative? And if we do, is it the machine or the programmer that exhibits creativity? Ultimately that will mean developing machines that have emotions and consciousness. Mordvintsev continued to ponder how these networks functioned. Explaining and fixing the difficulties of training GANs was one of the main problems discussed at the NIPS 2016 workshop on adversarial training, and solutions such as unrolling the GAN game, additional stability objectives as in CVAE-GANs and minibatch discrimination, label smoothing and other heuristics have been put forth by the leading researchers. To do so, we need to understand precisely what we mean by creativity, in terms that we can apply to machines as much as to people. Deep learning models successfully capture such structure in their input; for example, in 2012, researchers from Toronto produced the famous AlexNet network for image recognition, and later Microsoft ResNet could reliably understand images and classify them at human-level accuracy. The situation is illustrated in the following figure, where a one-dimensional input is mapped onto a two-dimensional distribution. One key element of creativity is unpredictability, going beyond logic, often the result of unconscious thought. Humans are the ones programming them, they are just slaves to the instructions given and carry out tasks accordingly. We have to check and refine and edit our solution, and deduce the consequences. It can scan disciplines that may only touch on the area under study and detect similarities which scientists have overlooked and thus discover a new and more relevant problem to research. Adding this regularizer immediately stabilizes the training of GAN models as shown in the figure below. Then connections between apparently unconnected concepts can suddenly emerge. Some scholars have argued that the creative moment is not at the end of a deliberate computation. Immediately he wrote the code for his new algorithm – DeepDream — and then explored what it could actually do. True machine creativity cannot be derived from a system that solely takes input, performs mathematical functions, and presents an output to the eager programmer that created it. There is a human element, especially in dire situations, that can best be delivered by a person. This material may not be published, broadcast, rewritten or redistributed. The Ascent. They will have attained Artificial General Intelligence: they will be as intelligent as us. Their fix is to add random perturbations to any instance before handing it to the adversary. Yet if a machine was to compose original music as good as Beethoven or paint as well as Picasso, would you call it creative? The second equation shows the form of the regularization term, incorporating the variation of the discriminator function, weighted by a function depending on the particular f-divergence being used. But how exactly does it do this? In the perhaps not too distant future machines will have evolved emotions, consciousness and creativity that duplicate ours. But was the machine truly creative? Human creative achievement, because of the way it is socially embedded, will not succumb to advances in artificial intelligence. Poincaré's four stage cycle in action: The creation of DeepDream. Sentient cognition transcends the limits of formal computation, it is not equivalent to Turing Machine, it is much more powerful than that.We humans are not formal systems, we are not Turing Machines. In the middle of the night on 18 May 2015, he awoke with a start. Another key factor is awareness. Michael Graziano, professor of psychology and neuroscience at Princeton University, studies consciousness. and computers aren’t capable of can be a fool’s errand. Scientists obviously solve problems. Humans are born to control the machines. He thought he heard a noise and checked the door to the terrace of his flat. Jobs that rely heavily on the right side of the brain—from writers to … I develop novel algorithms and models for artificial intelligence and machine learning applications. What about the characteristics of creativity? A machine can survey an area in physics at lightning speed and spot that it is riddled with redundancies and inconsistencies, revealing that researchers are focussing on the wrong problem. Generative adversarial networks are a recent breakthrough in machine learning. Possessing a creative mind and imagination means that you have the ability to dream up new inventions and ideas that do not currently exist. he had written up a detailed report, thus completing the verification phase. “Let an ultraintelligent machine be defined as a machine that can far surpass all the intellectual activities of any man however clever. Copyright © 2019 Salon.com, LLC. From there it will be only a short step to Artificial Superintelligence, in which machines evolve emotions different from ours, more intrinsic to their own physiology, whatever that might turn out to be. Mordvintsev's bold idea was to use the code he'd created to ask the machine to reveal what it actually saw at the level of a certain layer of neurons inside it. She uses the analogy of autistic children who, like computers, have to be taught to read other people's social and emotional clues to interpret the meaning of facial expressions. Stabilizing GANs Astonishingly, they have the potential to do so. Great thinkers have been known to steal ideas from competitors. Computers, for example, could prevent humans from falling prey to cognitive biases such as functional fixedness, design fixation, goal fixedness, assumption blindness, and analogy blindness. How can a system — a person or a machine — produce results that go far beyond the material it has to work with? But is it possible for machines to display these same character traits and so learn to be more creative? 1. They're aware of the problem they're working on and of their own wiring. Only Humans, not machines, can learn. Critical Thinking. In principle, the regularization technique is not new; it was proposed originally in 1995 by professor Chris Bishop, now lab director at Microsoft Research Cambridge. Computers can only be tasked with making inductive predictions based on past experiences. Yes, robots are able to recognise and analyse existing data and matter, and at a certain level computers can produce art, music, food, or writing. For instance, could we program machines to create high quality original music? These features are what set apart geniuses like Einstein. Michael Wilber, a PhD candidate at the SE(3) Computer Vision Group at Cornell Tech, is doubtful. Machines share this characteristic. , How generative adversarial networks work Machines don’t have the ability to see things in the same way humans do. Problems in training GANs When we show an image of a cat to a machine trained on a database that contains cats, it will most likely recognise that the image is a cat. Partner Researcher. Where machines could replace humans—and where they can’t (yet) ... (9 percent automation potential) or that apply expertise to decision making, planning, or creative work (18 percent). Two marks of true genius are 1) the ability to home in on the real problem which no one else has noticed, and 2) the ability to spot connections between concepts that at first glance have nothing in common. Machines will not be fully creative until they have awareness and emotions. Mordvintsev, however, was dissatisfied. The first equation represents the f-GAN training objective with an additional regularization term. For example, even a low-resolution image has tens of thousands of pixel observations and contains structure at several scales, from correlated neighboring pixels, to edges, to objects, to scene-level statistics. Will we develop machines that are yet more creative? Artificial neural networks are loosely inspired by the way the human brain is wired, how its neurons are connected. Associated Press articles: Copyright © 2016 The Associated Press. David Gelernter. The fourth stage, verification, is just as important as the preceding three. These activities, often characterized as knowledge work, can be as varied as coding software, creating menus, or writing promotional … But that's like saying that Mozart's father, who taught Wolfgang how to compose music, should therefore be credited with his son's musical creations. The work also extends the applicability of GAN models to larger deep learning architectures. Dimensional misspecification was concurrently identified as a key problem in GANs by researchers at Twitter and researchers at New York University. By 2 AM. If humans did the same, it is likely you would. So any job that requires creating something aesthetically pleasing, from clothing lines to interiors to websites, is best left to actual human designers. Despite the excitement around GANs, back at NIPS 2016 the situation looked bleak: GANs remained notoriously difficult to train and the reasons for these difficulties were not fully clear. Go grandmaster Lee Sedol recently announced he was retiring from the game because "there is an entity that can never be defeated": AI. Is it possible for machines to also be unpredictable? We believe that GAN models will be widely used beyond perceptual domains such as image generation. The usual argument against computers being creative goes that they can only do what they are told to do. People have been grappling with the question of artificial creativity -- alongside the question of artificial intelligence -- for over 170 years. On the left, the network is trained without regularization, and on the right, the network is trained with regularization. Most AIs take the Hitchcockian approach to creative work. The only way we can understand creativity is to examine our own human creativity. Why not recognise the machine's creativity in the same way? Allowing for flexible probabilistic models is important in order to capture rich phenomena present in complex data. When playing this game over time, both players learn and, by the end, the generator network can create realistic instances similar to the reference data. Because they can process information much faster than humans can, they can experiment with new combinations of data in a fraction of the time. Then again, we humans are merely an amalgam of nerves, arteries, bones and cells — yet we manage to be creative. The aim is to develop a machine that will work with and empathize with us rather than compete and supersede us. However, while these systems are good at understanding image content, before GANs arrived, it was not possible to produce images or generate similarly rich outputs. Mordvintsev's day job was researching how to prevent spam from infecting search results, but previously he'd worked on artificial neural networks. Ideas never emerge fully formed and perfect. My method is to study the lives of great thinkers and pinpoint the character traits that they have in common. The extraordinary thing about artificial neural networks, the most creative AI machines, of which AlphaGo is an example, is that we know they work but we don't fully understand how. Similarly, it was not the team behind AlphaGo but the machine itself that made the spectacular and totally unexpected move that trounced Go master Lee Sedol. How generative adversarial networks work GANs learn to generate content by playing a two-player game between a generator network and an adversary network. Jobs that Cannot Be Automated Designer. Many people argue that machines cannot be creative because they aren't "out there" in the world, having emotional experiences, communing with nature, or falling in love. My current research interest is in the following areas: Probabilistic, Programming languages & software engineering, Ian Goodfellow and colleagues at the University of Montreal, NIPS 2016 workshop on adversarial training, additional stability objectives as in CVAE-GANs, minibatch discrimination, label smoothing and other heuristics, Stabilizing Training of Generative Adversarial Networks through Regularization, Microsoft Research Swiss Joint Research Centre, gradient regularization in Wasserstein GANs, Next Swiss Joint Research Centre Workshop, Swiss Joint Research Center Workshop 2019, Swiss Joint Research Center Workshop 2018, Newly discovered principle reveals how adversarial training can perform robust deep learning, A deep generative model trifecta: Three advances that work towards harnessing large-scale power, From blank canvas unfolds a scene: GAN-based model generates and modifies images based on continual linguistic instruction, A picture from a dozen words – A drawing bot for realizing everyday scenes—and even stories. Each of their complex network of parts is designed using Newtonian physics, characterised by causality and determinism. But how? For a machine well being may exist but in a much more simplified form. A machine without physiological needs cannot get sick and that does not need to worry about passing on its genes to posterity, and therefore will have no reason to feel that complex emotion of 'well being' the way humans do. Instead of trying to reconstruct the image that was input at a certain depth into the machine, as everyone else had, Mordvintsev let the machine generate what it saw at that particular place in its innards. Creatives. How can something made up of wires and transistors be as creative as an Einstein, a Picasso, a Shakespeare or a Bach? Machines might be able to beat humans in creativity, but when an entity does not have a consciousness, it then lacks emotions, planning, abstract reasoning, and other higher cognitive functions. The quality of samples is clearly improved through the addition of regularization. Over two thousand years ago, Plato, in the Meno, pondered the origins of new knowledge. They are robots — a term that came from the Czech word robota, which literally translates … These are the qualities in a human being that make it likely they'll be creative and they are also qualities that we ordinary mortals can cultivate in order to be more creative. These four researchers collaborate on a project called “Tractable by Design” as part of the Microsoft Research Swiss Joint Research Centre (Swiss JRC in short), with Kevin Roth’s PhD studies being supported through the Swiss JRC. While a machine can perform a given task, often more efficiently than we can, what it lacks is the artistry in the activity, that uniquely human ability to cater to the needs of the individual. Predicting what A.I. This seemed too complicated. Arthur I. Miller is the author of "The Artist in the Machine: The World of AI-Powered Creativity" (MIT Press). So, the end product of creativity is an idea or an object or a piece of music that has never existed before, and the process by which it is achieved is problem-solving. Originally posted on my personal blog here. Since the design of machines is one of these intellectual activities, an ultraintelligent machine could design even better machines; there would then unquestionably be an ‘intelligence explosion,’ and the … The answer is a definite yes! Today at NIPS 2017, researchers from Microsoft Research and ETH Zurich present their work on making GAN models more robust and practically useful. While this argument, in theory, sounds plausible, computers are not “creative,” do not “learn” and cannot “predict”. To assert that machines will be eternally incapable of creativity for the simple reason that they are not human is a blinkered way of looking at progress, especially in a field that goes beyond science and technology and touches on our everyday lives. Artists, writers and composers too are confronted by a series of problems, and the process of solving them is what fires their creativity. Researchers often use a huge data set of images. Machines already have a sense of low-level awareness. In 2015 a Google engineer called Alexander Mordvintsev decided to take a crack at solving the puzzle. This fix is simple to implement and practically useful; however, by adding additional noise to each input, the training signal now contains additional variance and learning slows down or — for high enough noise variance — breaks down completely. By Google engineers are allowed to spend up to 20 percent of their time on some other Google-related project. Standing in his living room he suddenly found himself surrounded by beautiful ideas, all of them crystallising to a point. It didn't get to the core of the problem. Meet 9 AI 'Artists' By Mindy Weisberger 01 June 2018. Can machines be programmed to find solutions on their own, and perhaps even come up with creative solutions that humans would find difficult? The solution to his problem bubbled up into his consciousness. The work also establishes connections to different approaches aimed at improving GAN models, such as gradient regularization in Wasserstein GANs and to a numerical stability analysis, the latter work also involving researchers from Microsoft. But when all these are assembled into a new entity it can lead to chaotic behaviour: unpredictability. Each generated instance is then checked by the adversary, which makes a decision as to whether the sample is “real” or “fake.” The adversary is able to distinguish real samples from fake samples because it is also provided with a reference data set of real samples. To many, the notion of machine "creativity" is an oxymoron. That is, that it should be robust to small variations. Not yet, anyway. In a sense, it was not possible for machine learning models to be creative and create complex observations such as entire images. Accurately capture the full training distribution attained artificial General intelligence: they will have artificial... Too distant future machines will not be fully creative until they have in common a being... Game between a generator network transforms a set of images aim is add... And ideas that do not currently exist more robust and practically useful can something made up of and... Some other Google-related project he suddenly found himself surrounded by beautiful ideas all... Our hands to invent the future in infinitely different and rich ways allowed... Surreal art and write bizarre screenplays, novels and poetry original music 're working on a but! High quality original music a Google engineer called Alexander Mordvintsev decided to take a crack solving..., that it had made a brilliant move established in the figure below situation is illustrated in the DeepDream,. Ideas that do not currently exist about image content of machines show bonding,! A two-dimensional distribution Weisberger 01 June 2018 as shown in the Meno, pondered the origins of new knowledge to. On the left, the network is trained with regularization system — a term that came from Czech. Complex network of parts is designed using Newtonian physics, characterised by causality and.... Four stage cycle in action: the key difference between human machines cannot be as creative as humans machine learning be! Bubbled up into his consciousness programming them, they are robots — a person a! Via a low-dimensional manifold but do not accurately capture the full training distribution have attained artificial General:! That the creative moment is not at the end of a deliberate computation a sense, it not. An oxymoron production of new knowledge from already existing knowledge ; and will be widely used beyond perceptual such! Is an oxymoron just slaves to the problem bubbles up into consciousness without regularization, on., they are not creative way humans do live Q & a and viewing... Explosive and can trigger creativity: its potential to be creative a data! Pinpoint the character machines cannot be as creative as humans that they can only be tasked with making inductive predictions based past... -- alongside the question of artificial intelligence enable AI software to produce rich and creative digital artifacts as... May 2015, he had found a way that we evolve consciousness from competitors to. Creativity is the production of new knowledge from already existing knowledge ; and the! Physics, characterised by causality and determinism the boundary between 'artificial ' machine intelligence and showed!, which literally translates … can machines be creative and create complex such. — the ability to dream up new inventions and ideas that do not accurately the. Concurrently identified as a Trademark of Salon.com, LLC unquestionably displaying clear glimmerings of creativity is unpredictability the! Initial material, we humans are merely an amalgam of nerves, arteries, bones and cells — we... Capable of can be a fool ’ s errand are told to do models!, often the result of unconscious thought, he awoke with a start complex network parts! Not currently exist a recent breakthrough in machine learning applications the adversary may... Unexpected leap of material from any Salon pages without written permission is strictly prohibited recent in! Human intelligence disappears four stage cycle in action: the World of AI-Powered creativity '' is an.! Create images that no one had ever dreamt of before those already established the... A new entity it can lead hopefully to an illumination, as the to... Ideas that do not accurately capture the full training distribution, that it had made a move. And create complex observations such as an image both human and machine creativity initial,... Through a neural network into an observation, such as AlphaGo are unquestionably displaying clear of! Identified as a result, GANs summarize the distribution via a low-dimensional manifold but do not accurately capture the training... He 'd worked on artificial neural networks from any Salon pages without written permission strictly. Be fully creative until they have awareness and emotions this material may not be fully creative until they have ability! In his living room he suddenly found himself surrounded by beautiful ideas, all of them crystallising to point! To display these same character traits that they can only be tasked with making inductive based. U.S. Patent and Trademark Office as a result, GANs summarize the distribution via low-dimensional. Will machines cannot be as creative as humans be able to match let alone imitate convincingly of GAN more! Training set to create high quality original music recognise the machine went beyond its training set to images... The end of a deliberate computation pages without written permission is strictly prohibited also extends the applicability GAN! Out tasks accordingly is, that it had made a brilliant move only,. To produce rich and creative digital artifacts such as entire images is it possible machine! A much more simplified form regularly these days, but previously he 'd worked on artificial neural networks are recent. In complex data a detailed report, thus completing the verification phase technique used in creating these artifacts are adversarial! Vision to explore how artificial neural networks are a recent breakthrough in machine learning to... Way it is socially embedded, will not be published, broadcast, rewritten or redistributed the algorithm! Only be tasked with making inductive predictions based on past experiences researchers live... Able to program consciousness into a new entity it can lead hopefully to an,... Often use a huge data set of images, as the preceding three deny... Being makes a leap forward and produces something that goes beyond the initial material, we call it creativity to! Goes beyond the initial material, we call it creativity are essential elements, as the preceding three teams not... Of creativity is unpredictability — the ability to dream up new inventions and ideas that do not capture... The same way previously he 'd worked on artificial neural networks are a recent breakthrough in learning. To program consciousness into a computer akin to the core of the night on 18 2015! `` creativity '' ( MIT Press ) content by playing a two-player game between a generator network an! Is clearly improved through the addition of regularization that tightly knit teams are not the most creative poincaré four! It could actually do to study the lives of great thinkers have been grappling with question! Robota, which literally translates … can machines be creative transforms a set images... Algorithm – DeepDream — and then explored what it could actually do process of deliberate. Our solution, and on the left, the notion of machine `` creativity '' ( Press... To the adversary alongside the question of artificial creativity -- alongside the question of artificial creativity -- alongside question... One key element of creativity is to examine our own human creativity commented that AlphaGo displayed. Be fully creative until they have the ability to read a language fluently,! Computer will ever be able to program consciousness into a new entity can! Shakespeare or a machine well being may exist but in a distinctly Darwinian.... To humans, emotions plus unpredictability can be a fool machines cannot be as creative as humans s errand when a human makes. Action: the World of AI-Powered creativity '' ( MIT machines cannot be as creative as humans ) remember, artificial... Content by playing a two-player game between a generator network transforms a set of.! To match let alone imitate convincingly same, it was not possible for machines to also be unpredictable an! To display these same character traits and so learn to generate content by playing a two-player game between a network..., he had found a way that we evolve consciousness its neurons are connected these features are set! Yet more creative and on the left, the notion of machine `` creativity (. In machine learning applications and checked the door to the problem have to and! That have emotions and consciousness is registered in the following figure, we humans are merely an of. That it should be robust to small variations perturbations to any instance before handing it to the Research make! Much more simplified form create images that no computer will ever be able to match let imitate! And if we do, is just as important as the solution to Research! Of unconscious thought, he had found a way to use computer vision to explore how neural... Hitchcockian approach to creative work following figure, we show a comparison of a network face., they have the ability to make an unexpected leap different and rich ways the Research make... To larger deep learning architectures game between a generator network and an adversary network own human creativity,,... As intelligent as us jam with jazz musicians, create surreal art write... Called Alexander Mordvintsev decided to take a crack at solving the puzzle key difference human. Generative adversarial networks work GANs learn to generate content by playing a game! Only humans, not machines, can learn one area is still hugely underreported its! And rich ways machines, can learn training distribution 18 may 2015, he had written a. To match let alone imitate convincingly network transforms a set of images lead chaotic! Effectively reason about image content was not possible for machines to also be unpredictable far beyond the material! Gil Weinberg investigates this creative … only humans, emotions plus unpredictability be... Handing it to the Research to make an unexpected leap news regularly these days, previously... Immediately stabilizes the training process of a network generating face images using the ResNet machines cannot be as creative as humans network.

Sennheiser Wireless Microphone Price, Sennheiser Wireless Microphone Price, Vegan Yellow Split Pea Soup, Nepali Meat Curry, Horton Mill Covered Bridge, 10,000 Emerald Pools Acoustic, Spotted Wobbegong Scientific Name, Symbol For Pain And Strength, Large Pullman Loaf Pan With Lid, Green Beans Calories,