Loading activity…
@mikeknoop
co-founder @ndea and @zapier @arcprize
AI
ARC Prize plans to name private-leaderboard winners December 4, with paper submissions open another week as of November 3.
AI
The reported MATH-500 score rose from 7.9% to 48.1% with the same sampling settings.
AI
ARC Prize says plotting model scores reveals the pattern; another post calls evaluation science nascent despite its importance.
AI
A poster argues that judgment and taste will remain hard for AI to master across domains and over time.
AI
A post calls GDM’s results impressive and notes a “backwards bend” in action efficiency.
AI
Goodfire researcher Nick Cammarata argues that modern tools reveal most neurons as legible in models up to trillions of parameters.
AI
A commentator disputes AI 2027's human-response forecasts, arguing labs accelerate when perceived risk falls and act when it rises.
AI
A user says coders no longer read code and mathematicians solve decade-plus open problems, while calling crypto increasingly fragile.
AI
A post recommends brand awareness and SEO, public evaluation scores, and inclusion as a specific post-training skill.
AI
An October 9 post says 'we're reaching' those scores now, without giving any figures.
AI
National Compute says .edu, .mil and .gov email holders can sign in, while others can join a waitlist.
AI
OpenAI says it drew on an independent mathematics advisory group’s advice about how to release the results.
AI
Its team claims activation perturbations can approach backprop, but an outside researcher questions the evidence.
AI
A commenter calls the verification notable, saying the model is about as big as can fit within the competition’s official runtime.
AI
Mike Knoop says the result came under strict Kaggle compute limits with no internet access, on what he describes as a much harder reasoning challenge.
AI
One commenter sees a maturing LLM market opening up useful new ways to package parts of the technology. Another says logprobs had become “fairly useless” before they were removed.
AI
A user quotes OpenAI as claiming more than 100 resolved open problems, plus the Navier–Stokes Millennium Prize problem. Separately, OpenAI announced an independent mathematics advisory group.
AI
Probably’s creator says Jev makes the decisions, a large language model does the writing, and a small program ties it together.
AI
The post argues that decisions cannot be better than optimal, making data acquisition and lower costs the most plausible near-term opportunities for recursive self-improvement—AI improving itself.
AI
The user argues that AI can help rule out insufficient searching as an innovation bottleneck, allowing a shift toward developing new prerequisite ideas.
