The Artist in the Machine: The World of AI-Powered Creativity

by Arthur I. Miller

An authority on creativity introduces us to AI-powered computers that are creating art, literature, and music that may well surpass the creations of humans.

Today's computers are composing music that sounds "more Bach than Bach," turning photographs into paintings in the style of Van Gogh's Starry Night, and even writing screenplays. But are computers truly creative--or are they merely tools to be used by musicians, artists, and writers? In this book, Arthur I. Miller takes us on a tour of creativity in the age of machines.

Miller, an authority on creativity, identifies the key factors essential to the creative process, from "the need for introspection" to "the ability to discover the key problem." He talks to people on the cutting edge of artificial intelligence, encountering computers that mimic the brain and machines that have defeated champions in chess, Jeopardy!, and Go. In the central part of the book, Miller explores the riches of computer-created art, introducing us to artists and computer scientists who have, among much else, unleashed an artificial neural network to create a nightmarish, multi-eyed dog-cat; taught AI to imagine; developed a robot that paints; created algorithms for poetry; and produced the world's first computer-composed musical, Beyond the Fence, staged by Android Lloyd Webber and friends.

But, Miller writes, in order to be truly creative, machines will need to step into the world. He probes the nature of consciousness and speaks to researchers trying to develop emotions and consciousness in computers. Miller argues that computers can already be as creative as humans--and someday will surpass us. But this is not a dystopian account; Miller celebrates the creative possibilities of artificial intelligence in art, music, and literature.

Genres: Technology, Art, Artificial Intelligence, Nonfiction, Science, Art Design, Design

432 pages, Hardcover

First published September 6, 2019

Reviews

[Emre Sevinç · Rating 4 out of 5]

As a sort of compilation of the state-of-the-art in artificial intelligence based creativity in music, literature, theater, humor, painting, graphics and animation, I can easily award this book five stars. It not only does a very decent job of describing the various interesting artificial intelligence systems, but also portrays the people that "created" those systems (should we call them meta-creators, one wonders!). I find it satisfying to learn a bit about the background of those technology experts and artists, as well as their different approaches to involving software and sometimes cyber-physical systems in the creative process in very different domains.

There is also a good deal of discussion about definitions and various aspects of creativity, both from the perspective of human creativity as we know it, and how fundamental terms and concepts can be, or whether, related to computer-based creativity in meaningful ways. I also liked those parts of the book, forcing me to go back to the master creators and the works I appreciate in music, literature, and fine arts.

I'm afraid I can't give much praise for the final few chapters that are on the topic of consciousness: I found it so shallow that I can claim removing those parts would even make the book better. And I'm surprised because I'd expect an emeritus professor of philosophy of science to be a little more nuanced. Any informed readers who's spent more than a few months on the thorny topic of consciousness knows the pitfalls and the need to tread lightly and very carefully in this domain. Long story short, that part is probably 2-star material. (Maybe I should go back to Thomas Metzinger and Saffet Murat Tura for a better treatment!)

Long story short, if you are curious about the most interesting, modern and captivating examples of artificial intelligence based creativity in various fields of arts, this book is a very a good summary and starting point. If you're after a critical and philosophical, as well as psychological investigation with respect to artificial intelligence creativity, it'll probably fall short of your expectations.

PS: You might also like the following short blog entry that provides a few interesting examples in artificial intelligence based music: "Generative Deep Learning and Bach, a Good Fit?"

PPS: It was also nice to see artists and technologists from Istanbul make an appearance in this book (one of them from the university where I worked almost a decade :)

[Silas · Rating 4 out of 5]

This was a joy to read. It doesn't get very technical on the AI's inner-workings; you'll need to dive into a machine learning textbook for that. It's instead a guided tour through a theme-park, watching just enough of each exhibit to get the gist of what the creator and its creations have accomplished. It'll definitely have you reading a few of its short chapters, then mulling it over with a blunt or two.

It would have done well as a documentary film. Thankfully, the book does include a link to a webpage full of audio/visual examples of the creative AI discussed in the book. But it wouldn't have killed them to include a few more full-color pages for at least the "AI as an Artist" section of the book.

[Humphrey Hawksley · Rating 5 out of 5]

Here is the most refreshing and lucid examination of artificial intelligence where most focus so far has been dystopian, in rigging elections or gaining battlefield supremacy. In a highly readable account, Arthur I Miller, turns his forensic eye to the upside of the artist robot. He explains how creativity, whether from a human brain or computer, follows a path of accumulating experience, learning or 'data'. Computing is now so advanced that robots can compose music more complex than Bach, crack jokes that have audiences reeling with laughter and weave magic with words to parallel the genius of Shakespeare. Of course, all this raises questions about using algorithmic code to replicate human emotion, but Miller's take is that we should look on the up side and not be afraid. Miller's own journey partly parallels that of a computer in the way he gained and processed experience (or data) data that led him into a world of creativity. He was born into a dysfunctional family in New York’s Bronx and fell through the local education system. His sanctuary was reading, libraries and one day he heard Tchaikovsky’s Fifth Symphony. He was blown away and has been fascinated in the process of creativity ever since. This is a great, accessible read on one of the most important topics of our time.

[Mizuki · Rating 5 out of 5]

It seems like a textbook covering major recent topics about AI and arts. I also enjoyed the author’s webpage containing color images and music/videos, I strongly recommend to access that page, especially if you’re reading on kindle.
https://www.artistinthemachine.net/

Personally I feel rather comfortable with the idea of ‘creative computer’, it helps expanding our cognitive world. As an amateur pianist, I also like music composed by computer, but I feel like that the actual ‘playing’ by robots are still far behind of human’s play, especially for the acoustic instruments. Anyway I’m thrilled for the (near/far) future of the arts by reading this book. I don’t think it would be a horrible world!

[Miles Wood · Rating 2 out of 5]

Although Miller provides a practical, readable, and intelligent account of the history, current state, and potential future of machine creativity, he fails to fulfill the expectations he establishes in the introductory chapters. Rather than illuminating a landscape that is transforming creativity as we know it, he reveals a series of technologies that more realistically serve as tools and partners to a new breed of artists. Furthermore, his assumption of computationalism and avoidance of other theories is misleading given our uncertainty of philosophy of mind. With that said, machine creativity is in its infancy and this book, alongside others, ought to play an important role in shaping the public’s appreciation for what is possible. The topic of creativity will likely become increasingly important to understand the complexity of machines we are building, and moreover to contemplate what role humans might play in a technology-driven future.

[Anders Sundell · Rating 2 out of 5]

Intressant ämne, men tradigt skriven, och med ytlig analys. Boken är som en serie Wired-artiklar, där olika kreatörer intervjuas om sina program, i princip utan kritiska frågor. Många liknelser och tankefigurer återkommer. Boken hade behövt redigeras hårdare.

Men den stora bristen är att Miller inte går på djupet med de viktiga frågeställningarna. Turingtestet refereras, men används inte. Poängen med testet är att man aldrig kan veta om en annan människa verkligen är intelligent, eller medveten. Om datorer kan efterlikna mänskligt beteende tillräckligt bra blir den rent epistemiologiskt intelligent på samma nivå som en människa.

Miller skriver också att "the great stumbling block is that computers cannot appreciate the art and music they themselves produce and are unaware of the quality of the moves they make in chess and Go." Det stämmer inte. Eftersom det finns ett tydligt mål med både Schack och Go - att vinna - kan varje drag utvärderas utifrån om det för datorn närmare eller längre ifrån det målet. I konst finns inget sådant mål: det finns ingen nyttofunktion att maximera.

Därför faller Millers återkommande fras att "i framtiden kommer datorer att framställa konst som inte liknar något vi är vana vid" platt. Varför skulle den göra det? Kanske, om AI uppstår, och datorerna blir medvetna. Men mer sannolikt på kort sikt är att AI används för att maximera mänsklig nöjdhet med konst. Till exempel genom att låta folk bedöma kvaliteten på musik genom lyssningar på Spotify. Fler lyssningar ger mer belöning till de neurala nätverken, vilket ger ledtrådar om hur musiken bör komponeras.

Faktum är att jag under vissa särskilt mallade och simpla delar funderade på om boken i sig var skriven av ett neuralt nätverk.

[Terry Koressel · Rating 2 out of 5]

I gave The Artist in the Machine only 2 stars, possibly due to a case of high expectations. I was so fascinated by the subject matter that I could not wait to start this book. In the end, though, I had trouble getting through The Artist in the Machine.

I appreciated Mr. Miller's in-depth explanations into the nature of creativity itself. This was fascinating and thought-provoking. But eventually the book evolved into separate historical accounts of artificial intelligence (AI) advances in numerous creative fields. At this point the subject matter became repetitive and dry....at least for me. One other problem with the book: Hardly a week goes by without a major new announcement of an AI breakthrough, new record or new approach. AI is advancing exponentially. Given the pace of innovation, The Artist in the Machine will have to be updated regularly in order to stay relevant.

[Lee Barry · Rating 3 out of 5]

If you've been following the use of AI and machine learning in the arts, this would be old news or things you've read in art magazines. In fact, I recall reading books in the mid-late 90s on these topics and not much has changed except the usual improvements in technology. The ideas are the same, at least to me: working with data sets and machine learning to generate abstract images or music. There's no "wow" in it anymore. Nonetheless, it is a very good overview of the sciart scene over the past 20-25 years. The book I recommend for a counterpoint is The Myth of Artificial Intelligence: Why Computers Can't Think the Way We Do. That changed my view of AI more generally.

[Michael Anderson · Rating 3 out of 5]

I almost liked this book. It’s primarily a survey of specific software research systems, mostly based on neural networks, that address a computer’s capability to independently create art, music, poetry, stories and play hard games such as chess and go. The variety of the research systems got to be repetitive after a while — who wrote it, what the program does, how it differs from other similar programs — but not so much was said about how the programs worked. I guess this disappointed me. The last chapter in the book, addressing “Can Computers Be Creative?”, is the best, but it’s way too short. All in all, this is a good book, but it could have been better.

[__void__ · Rating 3 out of 5]

Found the sections diving deep into the meaning of creativity and genius very interesting (both at the start and end of the book). Sections about GANs, pix2pix and image generation were also interesting (even more now in the age of Stable Diffusion). Sections of music, poetry and writing were not as interesting to read because the networks were not well developed (apart from maybe Music Transformers). All in all, a great book to review all the progress in AI Art Generation.

[Peter · Rating 2 out of 5]

The MIT Press has a habit of getting one of their favourite pundits to write outside their speciality. There results are never good. This book is a prime example.

Miller is a historian of science and has written good books on folks like Einstein, Poincaré, etc. But an "authority on creativity" he is not. This is obvious from the outset where he defines creativity as "Creativity is the production of new knowledge from already existing knowledge and is accomplished by problem solving" which doesn't align with any of the literature. (Anything that includes both 'novel' and 'useful' is generally ok. And Miller's inclusion of 'existing' and 'problem solving' is problematic.) And the book goes downhill from there.

The book is a wealth of technical details on things folk have built, and a void when it comes to creativity. It's not just wrong, it's confused. The sad thing is that there are some really good books on computational creativity. Miller even has a whole chapter on Margaret Boden's work (her "The Creative Mind" is very good) but proceeds to make a mess of it. He even paraphrases Boden's definition of creativity "the ability to come up with products—ideas or artefacts—that are new, surprising, and of value" which is quite reasonable. And then goes on to compare it with the hot takes from various computer scientists.

The core problem is that Miller appears to have done a little research and then approach this as a biography. There's lots of interviews and presenting of opinions, and very little making sense of it all. The relevant theory is either ignored or wrong, the many technical details are therefore largely irrelevant, and Miller's synthesis is confused. (Who knows where mess of chapter 2 "Seven Hallmarks of Creativity and Two Marks of Genius" came. It's not in the literature and it's not in the interviews.)

Ignore this book. Instead read Margaret Boden, "The Creative Mind".

[Artur Coelho · Rating 4 out of 5]

O autor está claramente fascinado pelas possibilidades criativas das máquinas, dos algoritmos, como entidades autónomas. Tão fascinado que talvez lhes atribua maiores capacidades criativas do que aquelas que realmente têm, com um entusiasmo e deslumbramento que talvez ainda não se repercuta (ou o venha sequer a fazer, futuramente) no corrente estado da inteligência artificial, que, recordemos, é inteligente num sentido muito restrito, por muito surpreendentes, cativantes ou impressionantes que sejam os seus resultados.

Taylor analisa, sem ir muito a fundo, o trabalho de artistas e investigadores contemporâneos que se dedicam a explorar estas tecnologias como ferramenta criativa. Em diferentes vertentes, desde as mais superficiais - usar GANs para gerar imagens e despertar o hype nos media, por exemplo, às que aprofundam a ideia de criação artística como algoritmia com métodos, passos, sequencias e variações, e que agora pode ser feita mediada por Inteligência Artificial, o que permite explorar novas vertentes e obter visões surpreendentes. Aqui, talvez a música seja a área onde isto se manifesta de forma mais profunda.

Há um intrigante elemento comum à maioria dos artistas/cientistas abordados neste livro: o seu caráter verdadeiramente multidisciplinar, o serem pessoas formadas em computação mas que de desde cedo também exploravam vertentes artísticas, desistindo ou mudando de áreas por perceberem que a sua intuição, de que as tecnologias de computação também poderiam ser ferramentas de expressão plástica, era demasiado estranha ou ainda não aceite. Esse lado seduziu-me, o de especialistas em IA que nunca perderam a ligação à pintura ou à música, e que agora usam as ferramentas algorítmicas para mesclar os mundos da computação e da arte.

[Fehreen Khan · Rating 5 out of 5]

An in-depth overview of whether a machine can truly be creative and optimism of it being a conscious but logical being in the future once humans become extinct. I enjoyed learning of the various arguments by many authors before such as John Searl, Magaret Boden’s 3 types of creativity and Picasso’s good artists borrow and great artists steal. The concept of creativity shows it’s okay to build on what has come before but it becomes an issue when it is done by a machine rather than a human due to the lack of effort produced by the machine in comparison to a human. The book focuses heavily on creativity and mentions art and music from before technology to the more modern alphago and the Magenta music and art generation software. A great book but I would have liked a more realistic end rather than machines living harmoniously in another planet which they will find due to their technological abilities and advanced intellect once humans are extinct. As will AI be able to survive without any data being fed and how long until creations of art, music are unable to evolve due to a lack of data or is the possibility of creativity endless even if solely produced by machines, a hard to imagine reality.

[Ryan Trauman · Rating 4 out of 5]

Miller offers an excellent historical survey of artificial intelligence projects over the last several decades. His descriptions and explanations are clear, even though the subject matter is incredibly complicated. He strikes an effective balance between technical descriptions of the processes and the hardware involved. I know you can't include everything in every book, but I would have liked to hear more about how AI will obtain and use a sense of aesthetics. It wasn't until the very end of the book that Miller address ideas such as computational emotion, motivation, and appreciation. At the three-quarter mark, I was beginning to feel like we are on the cusp of building machines as artists. But when it comes to the three topics mentioned above, it's as though we've made almost no progress at all. It's only when you stop expecting aesthetic quality or emotional engagement that machines start looking like artists. And that changes the nature of the question altogether.

Also, I think his references stop around 2017, so if you're looking for insights into ChatGPT, you're not going to get much out of this book.

[Olivia Hoagland · Rating 4 out of 5]

This book was EXCELLENT for my thesis and my general interest on how our mind works. The comparisons between our neural networks and artificial ones was fascinating and I truly believe that computers have their own form of consciousness and creativity that is beyond our understanding, just like ours is to them. Some of the examples about music and poetry weren't my cup of tea but overall it was an outstanding book to understand the future of AI in the arts.

[Laurel · Rating 4 out of 5]

I was blown away by the plethora of examples of machine creativity. As a creative myself, I was excited to see how machine learning can generate music, literature, visual art, and much more! Because the book is composed a more a survey of examples, however, I found it pretty dense to read through cover to cover.

[Roberta Westwood · Rating 5 out of 5]

Excellent!

An excellent background on AI from the perspective of creativity. I liked how the author used Bach, Einstein and Picasso as genius human touchpoints as the aspects of AI creativity were explored. Have listened to several books on this topic, but while there was some overlap, there was plenty of new content. Highly recommended.

[Ineke Vermeulen · Rating 4 out of 5]

The subject is very interesting and this book gives a nice overview of all kinds of art projects that use AI. The interesting subject compensated for the sometimes a bit monotonous chapters with descriptions of the projects and artists, that all had the same format (written by an AI perhaps?).

[Brendan]

Just read some of

[Raumkundschafter · Rating 2 out of 5]

Nice title but unfortunately the author does not manage to look beyond the nose of a Computational theory of mind which makes it a rather uncreative approach to the topic.

[Antoinette Gouveia · Rating 3 out of 5]

Was too technical for me. I enjoyed the ways AI is becoming creative but the science behind them, was over my head.

[Thad McIlroy · Rating 3 out of 5]

I'm very interested in the intersection of AI and creativity, and am finding this book fascinating.

[John · Rating 4 out of 5]

Thought provoking and a really good collection of AI art up to the early 2020s.

[Clo Willaerts · Rating 4 out of 5]

Overview of creative computers and algorithms from human artists’ point of view

Most books on the topic of AI vs creativity are written by AI experts. This one first defines creativity and genius, and then gives an interesting overview of AI + art projects of the pas few years.

[Devin · Rating 3 out of 5]

There are a lot of stimulating case studies in this book, but they are built on a weak, hyperindividualistic definition of “creativity” that has been repeatedly debunked. (See Sam Franklin’s “The Cult of Creativity.”) In truth, the lofty objectives of art and science are not as similar as the gloss Miller presents.

[AP · Rating 5 out of 5]

Miller is more interested in output than intention. Which is fine, but it does make you wonder: If creativity is just remixing data in surprising ways, does that mean predictive text is a tortured poet?

[Local reader · Rating 3 out of 5]

I would describe it as a 300 page scientific review of AI in arts, in layman terms. Not too in depth, not too too shallow either.