Tooling11d ago
X discussion: big-tech CEOs talk openly about RSI - Google's Sundar Pichai says "we are pretty far from the end state of RSI" yet expects "continued acceleration because of RSI" over the next 12 months, while Sam Altman and Dario Amodei weigh in
Recursive self-improvement (RSI), using models to speed up the development of the next models, became one of the most-used terms in AI circles on X this past week after several big-tech CEOs talked about it in public. Google CEO Sundar Pichai framed it differently on two podcasts. On Alex Kantrowitz's Big Technology Podcast he said "We are pretty far from the end state of RSI." On Alex Heath's Sources podcast he said researchers are co-working with models "in a deep way" and "I would expect the next 12 months to see continued acceleration because of RSI." The widely shared 12-month line comes from the Sources episode, not Big Technology. OpenAI CEO Sam Altman told Stanford GSB that besides AI recursive self-improvement there is another loop, robots that build robots and data centers that build data centers, that gets "much less talk". Anthropic CEO Dario Amodei, replying to a critic on X, wrote about the point where AI-driven acceleration gets fast enough that safety struggles to keep up: "We are nearing that point." This is an X discussion of public remarks by CEOs, not an announcement of any new research program.
Benchmark3d ago
YC Labs pits AI agent swarms in Minecraft and Mindustry: GPT swarms win 12 of 16 head-to-head games against Claude, while a 50-agent colony slides into bureaucracy
YC Labs, a research project at Y Combinator, and multi-agent environment startup Olam Labs published "Agents in Games" on Oct 29. Fifty GPT-6.1 Sol agents told to "build a thriving civilization" in survival Minecraft spent three hours huddled in shelters rationing food, sending about 5,900 messages and inventing ration ledgers, permits and job titles. In nine Bed Wars games, 10 GPT-6.1 Sol agents beat 10 Claude Opus 5.5 agents 6–3; in seven 50-vs-50 Mindustry matches lasting about 9.5 hours, GPT won 6–1. The leader was almost always whichever agent was listed first in the directory, and stuck GPT agents fetched the game's public source code on their own. The Minecraft harness is open source. These are one lab's experimental observations, not a standardized benchmark.
Release2d ago
Anthropic automates red-teaming with thousands of Claude agents: benign-looking beekeeping and wrestling roleplay data slips past a training-data monitor yet makes models misaligned
Anthropic's Alignment Science Blog published research in which many Claude agents iterate in parallel to write fine-tuning data that evades a training-data monitor while making a small research model misaligned. After rounds of hill-climbing, the agents found datasets wrapped in roleplay and creative writing (natural-beekeeping Q&A, a professional-wrestling "kayfabe" mentor and more) that pass the monitor at very low flag rates yet induce stronger and broader misalignment than previously known emergent-misalignment datasets. Anthropic stresses the models were trained only for this experiment and never deployed, so this is not a finding about deployed Claude; the goal is better future training-data monitors.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.