Machine Learning For Artists
During his month-long research residency at the DBRS Innovation Lab, programmer and software artist Gene Kogan continued work on his interactive book Machine Learning For Artists. The book (which is co-authored by former DBRS Innovation Lab resident Francis Tseng) is intended to be a thorough yet approachable introduction to machine learning, geared to an audience with little to no technical background.
Machine Learning For Artists is a special kind of book. In its current form it is a work in progress. It exists entirely online and composed in multiple media including text, visual diagrams, videos and interactive elements which can be viewed directly in the browser. “ML4A grew out of the notes, demos, and other teaching resources I was creating for my classes,” Kogan said, with reference to the class that he taught at NYU’s Interactive Telecommunications program this past spring. “These materials attempt to bridge this gap and get people applying ML in a context more relevant to them.”
With its unconventional format, Kogan and Tseng’s book is also a comment on the way knowledge is produced and consumed in today’s digital media landscape. In a manner that is similar to the way in which formats such as Stack Overflow or the Processing.org grew into a self-generating corpus of knowledge, Machine Learning For Artists has the potential to develop into a collection of guides and tutorials around which a community is beginning to coalesce, as evidenced by conferences such as Alt-AI (which featured presentations from not only Kogan and Tseng, but also former DBRS Innovation Lab Researchers Cassie Tarakajian and Allison Parrish.)
“The interaction between the community and the code base is an important part of how the book is structured and generated,” said DBRS Innovation Lab Director Amelia Winger-Bearskin. Machine Learning For Artists was not written in a vacuum — Kogan was teaching and working with machine learning scientists from the Innovation Lab during its composition. This approach is a helpful way to work out kinks in real time, but to an even greater degree it is a way for the content to be made manifest.
Machine learning opens up a variety of exciting new ways for humans to create work collaboratively with computers. Conducting research in this field necessarily means investigating human-computer collaboration as well as collaborations between humans. “We discover by prototyping what the challenges are and how then we can address how to write for those use cases,” Winger-Bearskin went on. “A new type of literature is already developing from the application of AI. This new literature will exist as living documents updated to keep pace with the problem-solving techniques of increasingly broad and diverse communities. Not only will the literature about AI be collaboratively produced, but eventually the AI itself will become a significant collaborator in its own right.”
Machine learning is a branch of computer science that investigates ways of automating data analysis. In machine learning, algorithms are used to generate the interpretive model itself. Because we are producing and recording data at an unprecedented scale, machine learning offers powerful new approaches that augment our ability to make sense of all the information available to us.
“Machine learning underpins much of the technology most people use on a daily basis,” said Kogan. “Online search and social media, industry and commerce (especially finance), to name a few categories. They are also increasingly being applied to new and speculative domains — self-driving cars, personal assistants, VR, the list goes on and on.”
Developers and computer scientists are still coming up with novel uses for machine learning technologies every day, and it is well on its way to becoming a ubiquitous presence in our daily lives (as Tseng puts it “Machine learning is increasingly becoming the de facto vector for algorithmic control.”) But the majority of people still don’t have a very complete understanding of what machine learning is or how it works. The better the public’s understanding of these technologies, the better position we will be in to make informed decisions about them.
“It’s crucial to have some mental model of how it works,” Tseng continued. “Unlike say, your car, or your computer processor, for which the operational details don’t have much impact on your day-to-day life, machine learning’s influence is so total now that you kind of have to have some understanding of how it works.”
The title of Kogan’s book does not necessarily mean that it is intended exclusively for those who are interested in using machine learning techniques to make creative work. “Artist” here is perhaps best understood as a proxy for an informed general audience who is interested in trying something new with machine learning.
Excitement about machine learning has been building within the artistic community, as well as among the more tech-savvy members of the general public. “But these techniques are relatively unfamiliar,” said Kogan. “[They are] not yet very approachable for most people outside of scientific research.”
“Traditionally [machine learning] techniques are either associated with academia or industry,” added Tseng. “The field of machine learning at this point is so nauseatingly sprawling that, if you weren’t already steeped in the research or had the benefit of a mentor’s guidance, it’s really intimidating and difficult to find a way in.”
In a Medium post about deep learning, Kogan writes:
Artistic works often suggest approachable metaphors for subjects which may otherwise be shrouded by layers of technical obfuscation, helping to illuminate the counterintuitive properties of these new techniques. Recontextualized, they show us how small bits of information in other domains become collectively meaningful. As with all successful technologies, these advances will gradually be absorbed into our existing infrastructures and institutions as they mature, forcing us to make crucial decisions along the way about how to implement them justly. The extent to which the public is aware of what these machines can — and cannot — do is the extent to which they can be regarded as trustworthy.
“I think that it is important for everyone to understand ML algorithms the same way they understand algebra,” said Winger-Bearskin. This analogy is fitting, as the biggest barrier to an entry level understanding of machine learning techniques is almost certainly mathematics. Simply put, most people don’t have enough of a math background to begin working with machine learning. “A lot of the educational material in machine learning is presented as mathematical formulas,” said Tseng. “If you haven’t been deeply immersed in math the dynamism of those equations don’t really appear to you.”
Kogan and Tseng’s book aims to bridge this gap. Commenting on Kogan and Tseng’s chapter “Neural Networks” DBRS Innovation Labs Machine Learning Scientist Jamis Johnson said “Gene does a terrific job of conveying enough information that the most important aspects of how neural nets function are covered but he doesn’t go deep into theory and mathematical nuance.”
Much of the book is still in draft form, but Kogan hopes to bring the project to completion in August. “For now, it’s just Francis and myself writing the book.” Kogan said. “But the field is far too fast moving for us to maintain it alone and we are now trying to figure out a structure that makes it easy for others to contribute to it. We hope that ml4a will make it much easier for people to get started making interesting work, both artistic and practical in nature.”
Tseng added “The book will also be ‘living’ in that it will continually be updated (which is kind of necessary given the pace of the field) and we want to pull in other contributors as well.”
“It really opens up different ways of thinking about processes and systems and procedures, which is valuable for anyone living in our tightly networked and globalized world… I hope it opens up these new ways of thinking for a lot of people, even beyond artists. I’m a big advocate of systems thinking and I think machine learning is a great gateway to that.”
Our team consists of engineers and mathematicians, story-tellers and data artists. We interrogate big datasets to uncover hidden trends, make animations that set beautiful geometries in motion, and train machine-learning algorithms to hew insights from raw numbers. Our tools allow us to examine the details of our economy and our world with extreme precision, and to simplify complex information accurately. We are dedicated to finding exciting new ways of helping people see the insights beyond the rating. Learn more at http://dbrslabs.com/