**Foreword**

In 1843, the renowned mathematician Lady Ada Lovelace, often known as the world's first computer programmer, published her famous "Notes" (Lovelace 1843), and which, remarkably, included some insightful visions on computational art.

The "Analytical Engine" was a mechanical computer designed by the inventor Charles Babbage in 1837. It was only partially built in that time, but this did not stop Lovelace from designing programs for it and theorizing on its potential. She had the insight to see that it could perform beyond just acting on numbers to solve equations, but could also conduct symbolic manipulation to perform true general purpose computing. She also had the vision to foresee that the Analytical Engine could be used to compose generative music:

Supposing, for instance, that the fundamental relations of pitched sounds in the science of harmony and of musical composition were susceptible of such expression and adaptations, the engine might compose elaborate and scientific pieces of music of any degree of complexity or extent

or even that it could create generative imagery, "We may say most aptly, that the Analytical Engine *weaves algebraic patterns*, just as the Jacquard loom weaves flowers and leaves."

However, it is another of her rather controversial statements that I would like to recall:

The Analytical Engine has no pretensions whatever to *originate* anything. It can do whatever we *know how to order it* to perform. It can *follow* analysis; but it has no power of *anticipating* any analytical relations or truths. Its province is to assist us in making *available* what we are already acquainted with.

Almost two centuries later, we are still grappling with this statement and still trying to understand our relationship with the machine. Is it simply assisting us to make available what we are already acquainted with? Or can it originate anything? Can it anticipate any analytical relationships or truths?

We are not the first to revisit these questions. In 1950, over a century after Lovelace published her "Notes", the famed computer scientist Alan Turing addressed these topics in his seminal paper "Computing Machinery and Intelligence" (Turing 1950). He reframed the question in the context of surprise, asking whether a machine could ever surprise us. He added, "Machines take me by surprise with great frequency". However, his main proposition—echoing his collaborator Douglas Hartree—was that in order for a machine to really create something original, it should have a property that would not have been available to Lovelace or Babbage. That property, he concluded, was the ability to learn, "Instead of trying to produce a programme to simulate the adult mind, why not rather try to produce one which simulates the child's? If this were then subjected to an appropriate course of education one would obtain the adult brain."

This idea, now known as machine learning, is the very concept that underlies the recent surge in the fields known as artificial intelligence (AI) and AI art (though this latter label has not been adopted by everybody, including myself, for reasons which will soon become apparent).

The academic field of artificial intelligence is rooted in computer science but spans many other disciplines such as psychology, neuroscience, statistics, and philosophy. Since the 1950s, AI researchers have been thinking about the properties required to be able to create 'intelligent' machines, and they have been designing and building computational models of such systems.

For a similar amount of time, artists have been independently investigating the role of computers in art, creating many overlapping subgenres such as computational art, generative art, (new) media art, etc. As early as the 1960s, artists such as Harold Cohen were already engaging with AI (McCorduck 1990), and as early as the 1980s, artists were using machine learning (such as evolutionary algorithms) for their artwork.

However, just as the mainstream emergence of the internet in the 1990s produced a new era in computational art known as net art, the last few years mark a new era in computational art. As I previously mentioned, this new wave is colloquially referred to by some as AI art. Since it only refers to very specific recent technologies, it would probably be more accurate to describe it as (the much less catchy) 'deep learning art'.

Deep learning is a form of machine learning based on large artificial neural networks and massive amounts of data. The algorithms date back to the 1980s (or even earlier depending on who you ask), but only recently has the technology been broadly accessible and practically useful at scale. This is due to the vast increases in computing power that we have recently developed, and the massive amounts of data now available required to train these huge neural networks. The underlying reason for this recent explosion in deep learning is the political and social climate that requires, supports, and funds this research; the primary purveyors of deep learning research are also the primary purveyors of the data economy and mass surveillance—both state-sponsored and commercial—such as Google, Facebook, Amazon etc.

This situates deep learning, whether as a medium, tool, or subject matter, in a very unique position, and with very particular challenges and opportunities for artists.

Any topic surrounding big data immediately raises questions regarding privacy and ownership. On one end of the spectrum, we have organizations infringing on the rights of their customers by stealthily harvesting, selling, and otherwise exploiting their data. On the other end of the spectrum, we have individuals (such as artists) creating work using data belonging to others. In between, we have combinations that present more complicated ethical and legal challenges, such as the use of a neural network designed by one person or group, implemented by a second person, modified by a third person, trained by yet a fourth on data owned by a variety of individuals and collected by a fifth person, and using scripts written by a yet a sixth (true story). This highlights that our old concepts of ownership or even authorship, and our legal and economic systems built on these concepts, are becoming obsolete in these new digital ecosystems.

Given that deep learning can be thought of as a technology that attempts to extract meaningful information from vast amounts of big data, any progress in deep learning thus has the potential to impact any enterprise that is big-data-driven. Currently, practically all of our enterprises are big-data-driven, from physics, chemistry, and biology to finance, psychology, and even politics.

These deep learning algorithms are notoriously inscrutable. Often referred to as black boxes, they are incredibly difficult to decipher and meaningfully control or correct if they give undesired outcomes. As Alan Turing himself said in his 1950 paper, "An important feature of a learning machine is that its teacher will often be very largely ignorant of quite what is going on inside." This is also why they are so powerful. This is precisely what makes these big-data-driven learning algorithms both terribly exciting but also desperately terrifying. We are already seeing the unexpected negative consequences of for-profit companies using closed-source, closed-data, proprietary software to make critical decisions affecting thousands, if not millions or billions, of lives. We already have algorithms in use deciding which job adverts to show people, and not only that, but actually learning from current salary schedules to be sexist, by showing higher paid jobs to men only. We already have platforms, such as YouTube or Facebook, algorithmically spreading targeted misinformation and increasingly extreme propaganda. Extrapolating into the future, the potential damage these algorithms can inflict does not require a vivid dystopian imagination.

However, these are also the technologies that will allow us to see beyond what we would otherwise be capable of seeing, just as the telescope allowed Galileo, quite literally, to look at the sky and see the solar system in a new light. By helping us see and understand patterns and meaningful information in vast amounts of data, these big-data-driven learning algorithms can contribute to breakthroughs in many different fields, from helping us cure our most awful diseases, to early warnings for earthquakes or other natural disasters, to suggesting solutions for the ecological crisis we are facing. Some of these breakthroughs are likely to be unimaginable by our current norms, like houses that can grow organically or cars that can photosynthesize!

The future is not yet written; it is up to us to write it. Ultimately, technologies are not separate entities that are external to us. They are part of us. They are extensions of our bodies, extensions of our minds, and extensions of our values. They are embedded within us, and we are embedded within them—the users who use them, the researchers that develop them, the organizations that fund them —they cannot be totally separated from the motivations behind their development and the values of those who support and promote them.

As artists, we are in a unique position to help shape, or at least envision, potential futures. As artists, we try to trod untrodden paths. We try to imagine alternative realities and futures. We try to see and share different perspectives. We even try to feel those different perspectives and enable others to feel them as well. To quote Golan Levin, Professor of Electronic Art at Carnegie Mellon University, we try to create art that "comforts the afflicted or afflicts the comfortable".

I myself, as a computational artist practicing for almost two decades, have been thinking about such topics for many years, and I have decided to focus specifically on these big-data-driven learning systems, with meaningful human control, for my Ph.D. The authors of the essays in this book, collectively, have centuries of experience between them in computational art and creativity. We realize that we are together entering a new chapter in this area of inquiry, and, in the context of machines that can learn from vast amounts of data, that revisiting the relationship between artists and machines, and between humans and machines, is more urgently needed than ever.

#### **Memo Akten**

www.memo.tv

#### **References**

