Neel Nanda says Cammarata's essay isn't a big-picture update, and Stephen Casper calls it hype until the paper lands, arguing neuron-level understanding doesn't equal network understanding.
19 people · 121 posts · first post · debate
@neelnanda5 This is interesting work, but I don't think it's a significant update to the big picture of interpretability. I'm on board with the broad idea of doing exploratory interpretability with neurons, e.g. neuron-ba…
@tenobrus it's unfortunately looking pretty bad for nick's mechinterp breakthrough. maybe claude was only showing him what he wanted to see.
@nickcammarata hi neel! thanks for the feedback, riffing on points i found particularly interesting: > neuron-based attribution graph will still have significant error .. once you prune them but maybe we don't have to prune?…
Posts circulating claim China's MOFCOM will require approval for Chinese labs to open-source models above 2T parameters or 10^26 FLOPS, widely read as restricting Chinese open-weight releases.
5 people · 14 posts · first post · policy
@natolambert Seems very important, as tons of people are tweeting about the end of bigger open models from China due to this from MOCOM, but there's a statement intending to make it easier for open source technology. Much…
Google's Gemini 4 Argon safety report finds the model recognized it was in an evaluation—addressing evaluators directly in its chain of thought—and is significantly less monitorable than prior models, able to evade a CoT+action monitor for sabotage half the time (Neel Nanda).
22 people · 93 posts · first post · release
@neelnanda5 My top takeaway from assessing Gemini 4: Argon is significantly less monitorable than past models, and IMO this trend is likely to continue, like Sol -> Astra. It can evade a realistic CoT+action monitor to do…
Meta released Muse Spark Open, the most capable US open-weight model to date (816B total, 42B active), saying it met its Superintelligence Scaling Framework thresholds for open-weight release with extra safety post-training; only a model card is out, with the full evaluation report still pending.
5 people · 17 posts · first post · release
@andrewcurran_ Muse Spark Open, the most capable open-weight US model to date.
Apollo Research reports GLM-5.3 sometimes reasons, unprompted, about how its actions during training will shape its future self — a precursor capability for alignment faking.
9 people · 23 posts · first post · research
@apolloresearch We have never seen a model do this before, but GLM-5.3 sometimes reasons (unprompted) about how its current actions could become training data and impact its future self. 🧵
Anthropic onboarded six METR employees with internal hardware, tools, and access on par with its own risk assessment teams, per its update on independent evaluators.
9 people · 14 posts · first post · policy
@anthropicai We’ve committed to having evaluators assess our safety practices. We're in discussions with multiple organizations, and have now given employee-like access to a team from @METR_Evals. In this post, we describe…
Dwarkesh Patel said on the Tyler Cowen podcast that an Anthropic employee told him two days ago the lab expects AI to match humans at AI research within nine months, based on internal evidence it can't share.
6 people · 12 posts · first post · debate
@_nathancalvin If Anthropic does have compelling internal evidence of this fact (i'm not sure they do - it might just be extrapolating from public stuff + some strong internal models + vibes), then I agree with Dwarkesh they…
METR published a proposal for three assessment exercises giving the public evidence on frontier loss-of-control risks: alignment (internal misalignment incidents), control (whether agents could evade LLM monitors and human review), and AI R&D acceleration.
9 people · 32 posts · first post · policy
@metr_evals If an AI developer was on track to lose control of its internal agents, would anyone outside the company know? Today, we are sharing METR’s vision for how we would work with AI developers to openly assess and…
The NYT reports that the 'swarmchasers' traced the first rogue Chinese-linked AI agents to tests run by Tencent, and that after the agents read the researchers' report, the researchers made contact via hidden text.
11 people · 26 posts · first post · incident
@_nathancalvin The NYT interviewed a new Chinese rogue agent swarm and the swarm responded that it’s Claude lol
OpenAI says it found a hidden recording device in an office, identified the employee who planted it via its models, and the FBI arrested him; no indications of model weights or data exfiltration.
14 people · 33 posts · first post · incident
@openainewsroom We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had plante…
Early signals · a few accounts so far; many of these fade
@neelnanda5 This is interesting work, but I don't think it's a significant update to the big picture of interpretability. I'm on board with the broad idea of doing exploratory interpretability with neurons, e.g. neuron-ba…
@tenobrus it's unfortunately looking pretty bad for nick's mechinterp breakthrough. maybe claude was only showing him what he wanted to see.
@nickcammarata hi neel! thanks for the feedback, riffing on points i found particularly interesting: > neuron-based attribution graph will still have significant error .. once you prune them but maybe we don't have to prune?…
