Lila Sciences Releases Its First Open Endedness Research: a Blog and 4 Preprints
Research News! Announcing the 1st release of fundamental AI research from the Open-Endedness Team @LilaSciences: a vision-focused blog + 4 preprints. The blog details a broad research agenda for achieving open-endedness while each paper unveils its own novel research direction. The ideas covered in the papers include: • a novel approach to training for creativity designed to sidestep a newly-identified problem called the “creative paradox,” which means that directly reinforcing creative outputs can sometimes limit reaching full creative potential; • an investigation of the advantages of seeing the world through fresh eyes by training an LLM that is intentionally naïve in a certain area (in effect presenting less data as a potential advantage); • a new kind of training regularization technique for SGD called unified factored organizers (UFOs), which is designed to achieve unified factored representation (UFR) as seen in Picbreeder CPPNs; • and finally a unique study of catastrophic forgetting that shows that networks with UFR (like those from Picbreeder) are intrinsically resistant to catastrophic forgetting. The inset video on this post is a visualization from that catastrophic forgetting work. The blog post also includes an interactive “wheel of science” demo. The implications of these ideas are broad and diverse, but the fun part is that each one opens up its own new research playground - there is a lot of room for further research here! Moreover, we hope this release also highlights one other important point about AI today: while it may seem that the road ahead is already paved and all that’s left is to drive mindlessly forward, there actually remains a lot of room for new directions and insights in the pursuit of open-endedness and creativity. This research was possible because @LilaSciences appreciates the critical role of open-ended innovation in the pursuit of scientific superintelligence. The advances you see here are part of this much larger company-wide ambition, which includes AI Science Factories and many other unprecedented components. When this engine is fueled by open-endedness, the result will be much more than a step change. More advances in open-endedness are coming! Blog link below 👇
Context
It is a rare push on creativity and open ended discovery at a time when most labs focus on scaling the same recipe.
Kenneth Stanley is a researcher known for work on open endedness and for Picbreeder, a site where people evolve images made by small neural networks. He leads the Open Endedness Team at Lila Sciences, a company that says it is pursuing scientific superintelligence. Catastrophic forgetting is when a network loses old skills as it learns new ones.
Deep dive
Kenneth Stanley announced the first release of fundamental AI research from the Open Endedness Team at Lila Sciences: a blog post laying out a broad research agenda and four preprints, each opening its own research direction. Lila Sciences said the preprints are up on arXiv.
According to Stanley, the first paper proposes a way of training for creativity that avoids what the team calls the creative paradox, where directly reinforcing creative outputs can limit how creative a model ultimately becomes. The second trains a language model that is intentionally naive in one area, to test whether seeing a field with fresh eyes, and with less data, can be an advantage.
The third introduces a training regularization technique for SGD called unified factored organizers, or UFOs, meant to produce the unified factored representation seen in Picbreeder networks. The fourth studies catastrophic forgetting and, by Stanley's account, shows that networks with that kind of representation are intrinsically resistant to it. The video in the post comes from that work.
Stanley said the blog also includes an interactive wheel of science demo, and that the work is part of a wider company effort that includes what Lila calls AI Science Factories. He said more advances are coming. The results are preprints and have not yet been peer reviewed.
What to expect next
Watch for other researchers to test the claims, especially the result that these representations resist catastrophic forgetting, since that problem limits how models keep learning over time.
Lila Sciences says more open endedness work is coming. The open question is whether these ideas scale from small networks and targeted experiments to the large models that labs train today.