

Gregory Foster (@[email protected])
12.9K posts

@gregoryfoster
humbled by redwoods Founder + Citizen Observer @CannObserv tech @foojutsu words @entersection via @ConsumerReports @EFFAustin @gregoryfoster.bsky.social



"Move 37" is the word-of-day - it's when an AI, trained via the trial-and-error process of reinforcement learning, discovers actions that are new, surprising, and secretly brilliant even to expert humans. It is a magical, just slightly unnerving, emergent phenomenon only achievable by large-scale reinforcement learning. You can't get there by expert imitation. It's when AlphaGo played move 37 in Game 2 against Lee Sedol, a weird move that was estimated to only have 1 in 10,000 chance to be played by a human, but one that was creative and brilliant in retrospect, leading to a win in that game. We've seen Move 37 in a closed, game-like environment like Go, but with the latest crop of "thinking" LLM models (e.g. OpenAI-o1, DeepSeek-R1, Gemini 2.0 Flash Thinking), we are seeing the first very early glimmers of things like it in open world domains. The models discover, in the process of trying to solve many diverse math/code/etc. problems, strategies that resemble the internal monologue of humans, which are very hard (/impossible) to directly program into the models. I call these "cognitive strategies" - things like approaching a problem from different angles, trying out different ideas, finding analogies, backtracking, re-examining, etc. Weird as it sounds, it's plausible that LLMs can discover better ways of thinking, of solving problems, of connecting ideas across disciplines, and do so in a way we will find surprising, puzzling, but creative and brilliant in retrospect. It could get plenty weirder too - it's plausible (even likely, if it's done well) that the optimization invents its own language that is inscrutable to us, but that is more efficient or effective at problem solving. The weirdness of reinforcement learning is in principle unbounded. I don't think we've seen equivalents of Move 37 yet. I don't know what it will look like. I think we're still quite early and that there is a lot of work ahead, both engineering and research. But the technology feels on track to find them. youtube.com/watch?v=HT-UZk…

The dictator’s demise will be cause for celebration, though it will open up new dangers for the region and for U.S. interests there, writes Lawfare's Foreign Policy Editor @dbyman


What's the deal with Washington state?* (*assuming this isn't a we-counted-all-the-blue-ballots-first illusion?)




The Clark County Elections ballot drop box at the Fisher's Landing Transit Center was lit on fire this morning. Clark Co. Auditor Greg Kimsey says hundreds of ballots were inside at the time, the last pickup being 11AM Saturday. Full story here katu.com/news/local/van… @KATUNews

Here's a gift article of the Wall Street Journal investigation, which also confirms that Musk helped Russia kill civilians in Ukraine and privatizing government, like NASA's dependence on SpaceX, will always backfire and needs to stop wsj.com/world/russia/m…

So @every asked me to write about top level internet domain shenanigans. Come for the story of how a bunch of Slovenian academics stole the .yu domain. Stay for why the entire .io TLD might disappear due to a minor treaty between Britain and Mauritius. buff.ly/3zTcZyt

Christian nationalist pastor Joel Webbon calls for the public execution of women who falsely claim to have been sexually assaulted: "#MeToo would end real fast ... All you have to do is publicly execute a few women who have lied." peoplefor.org/rightwingwatch…