Loading activity…
@yonashav
ai resilience @foundationOAI. Past: @openai / @HarvardSEAS / @SchmidtFutures / @MIT_CSAIL. Tweets my own; on my head be it.
AI
Goodfire researcher Nick Cammarata argues that modern tools reveal most neurons as legible in models up to trillions of parameters.
AI
A post quotes David Sacks arguing that Claude’s training could increase the risk of frontier AI escaping human control.
AI
A post citing CBS says Philadelphia police disclosed the incident after Anthropic notified the department on October 7.
AI
The researchers say a Qwen student's letter-counting accuracy rose from 27.3% to 59.8% after training on unrelated answers.
AI
Usage policy update adds new rules and highlights the potential for misuse.
AI
A user quoted the post and wrote, “this looks dope.”
AI
Epoch AI fitted weekly figures through mid-August; Rohan Paul says the result reflects API list prices.
AI
The Atlantic argues the researcher's story went viral because readers want to believe his AI-doom claim.
AI
The Wall Street Journal reports DNI Jay Clayton will lead the group and effectively serve as AI czar.
AI
David Robinson argues OpenAI’s trial-and-error safety culture is inadequate for increasingly capable AI systems.
AI
CNBC reports the accord calls for internal monitoring, outside assessments and board oversight; Trump called it “morally binding.”
AI
A user argues that AI safety needs all three, comparing the choice to asking whether preventing theft requires laws, police or locks.
AI
A post urging people to join the AI Security Institute claims its team has found alignment issues, control monitor flaws and critical jailbreaks at “~all major AI companies.” It calls hiring the bottleneck.
AI
Replit argues the conversation should move beyond “AI safety” to who has access, what’s connected and how it’s secured.
AI
One post argues that race dynamics don’t exist outside AI’s frontier. Another says they exist across AI, with the most consequences at the frontier.
AI
Whitesides wants companies, independent experts and policymakers to establish clear red lines, arguing that AI is advancing faster than it can be understood and regulated.
AI
OpenAI researcher Dan Selsam argues in a personal statement that growing model awareness is undermining evaluation—and that more careful pacing of frontier AI alone will not adequately limit long-term risk.
AI
One post invokes “götterdämmerung” alongside “ai poaster,” without elaborating.
AI
A user calls for more attention to the group’s work, highlighting a standard covering embedded audits and members including METR, Transluce, RAND, AVERI, Princeton and SecureBio.
AI
The user’s experience-based comparison includes academic labs that measure AI capabilities and risks.
