Loading activity…
@rao2z
AI researcher & teacher @SCAI_ASU. Former President of @RealAAAI; Chair of @AAAS Sec T. Here to tweach #AI. YouTube Ch: http://bit.ly/38twrAV Bsky: rao2z
AI
Subbarao Kambhampati shared his University of Minnesota colloquium materials; his slides argue LLMs cannot plan autonomously.
AI
A user says they're excited for the future but grateful to have experienced research before large language models.
AI
NeurIPS says the papers will be presented across its conference locations in Sydney, Atlanta and Paris.
AI
Clark shared a livestream link and said his talk would cover the choices he believes people will need to make.
AI
A user asks for help choosing between what they see as two NeurIPS conferences in 2026.
AI
A user says some former Yochanites are apparently working to bring the two approaches together for robotics.
AI
A user gives November 12, 2026, as the effective date; another jokes about synths from the series Humans.
AI
A post argues that direct pixel-space training has an advantage because even good simulators are costly and lossy.
AI
A post quotes David Sacks arguing that Claude’s training could increase the risk of frontier AI escaping human control.
AI
The speaker said they planned to discuss LLM planning and reasoning in Saarbrücken and the “mythos” of thinking traces in Tübingen.
AI
The poster says they discussed the “questionable semantics” of reasoning tokens with WSJ journalist Christopher Mims and shared a paper cited in the article.
AI
The speaker says the trip is mainly for train travel, with plans to meet AI people along the way.
AI
A user says Claude spends a minute and a half thinking, then misrepresents what it thought about.
AI
A user jokes that adding LLM agents to Waymo cars’ workflows could let the company avoid responsibility for accidents.
AI
A researcher sharing new work says traces on iGSM, a benchmark proposed to highlight meaning in reasoning steps, lack end-user-interpretable semantics. Models trained with swapped or corrupted traces can still do well, they say.
AI
The New York Times reports that employees and security researchers said they cautioned OpenAI about safely testing its AI models and strengthening its corporate infrastructure, but the company did not listen.
AI
A post asks how much of the spring and summer run-up might have come from Meta paying full API prices for Fable traces to distill.
AI
A post alleges that AI labs spend millions, if not billions, experimenting at others' expense and falsely blame their negligence on 'agent intent.'
AI
One post compares AI discourse to climate change and evolution, arguing that Substack writers’ technical claims can be louder than scholarly voices.
AI
DAIR.AI describes a paper testing a two-model approach on 510 held-out preference pairs: accept confident verdicts from TypeSafe AI's JEV judge and send uncertain calls to GPT-6 Astra.
