My brother wrote a cool piece on AI in the music industry, and now they got us opening up AI conferences together. Join us for the pre-conference talk at ALGO-RHYTHMS on March 25. Details in comments. @dariusjagjag @cldpntr

Honored to announce that MOAL has accepted a Reflection into their Permanent Collection. To celebrate the event, will be releasing 20 Reflections from my personal collection with Iconic. Each will come with a robotic print as seen in this video..

Emerging Faces has come to be seen as one of the first instances of painting autonomously created by a robotic system, but this has never satisfied me because we don't really know what autonomy means, or what creation means for that matter. Perhaps art is always approaching a kind of recursive horizon of emergence, always embodying this threshold place.

Found this cool video from 2017 that shows how I was teaching robots to paint with Neural Networks (Mostly CNNs but you can also spot some early GANs in the background) - been doing this too long. youtube.com/watch?v=dkTjEi7O…

Emerging Faces Timelapse

Where does imagination come from? With all we know, we do not know this. For the viewer, an AI artwork can offer the experience of witnessing imagination imagine itself, like sending a spy into the threshold of emergence.

Chiang addresses a very real barrier that AI art will have to break through: the limitation of agency in using models that are constructed by others, very often by corporations developing commercial products to make the creation of art as easy as possible. The premise of his writing is that good art requires decisions, lots of decisions at every scale, and there is much truth in this. Do artists using AI processes have to sacrifice the integrity of nuanced decision-making? Artists can enrich the amount of decisions they're able to apply in such processes in a number of ways such as constructing narratives and manipulating outputs, but even these approaches have a hard limit as to how much influence they can have over the set of decisions, since the decision models themselves were designed by someone else. With these models becoming larger and larger, and more ubiquitous, we run the danger that all art will soon become homogeneous and represent the models themselves more than the artists. There are several artists who have posed different approaches to this dilemma, of whom Chiang seems to be unaware. Herndon and Dryhurst for their Whitney Biennale piece "xhairymutantx" used AI to make a conceptual statement about agency in the age of AI. They fine-tuned a diffusion model on images of Herndon and released it into the wild in order to interrogate whether we get to choose how we are represented on the AI substrate. This is not only art, but art that is vital to our times. Harold Cohen, with whom I was lucky enough to discuss this at length, made multiple AI systems to create art. Interestingly, he disliked machine learning as he thought that it was more of a filter that merely transformed one set of digital images into another at the press of a button. In this he probably would have agreed with Chiang. But he was more interested in creating systems that made actual decisions to create art from nothing. He wanted his software, AARON, to start with nothing but a set of rules, and create imagery from those rules. In essence he was interested in AI sculpting decision-making processes, perhaps of the type that Chiang is bemoaning the absence of. Does Chiang even realize that there are many artist focused on this kind of generative AI? In order to address the parts of Chiang's critique that I consider to be valid, I'm making a bunch of tools to help artists create AI models where their own decisions are key and will be releasing them over the next year. And I'm always around to help anyone that might be interested in pursuing this route.

When I make these, its always with a sitting where I talk the subjects ear off about cryptoArt, sometimes for hours, so the pleasure is mine.