Ben Levinstein

Member of Technical Staff at @AnthropicAI & Philosophy prof at Illinois interested in AI alignment, epistemology, and decision theory.

[Ben Levinstein @ben_levinstein]

Logging onto twitter for the first time in quite a while, only to say: We're on the right side of this dispute.

[Anthropic @AnthropicAI]

A statement on the comments from Secretary of War Pete Hegseth.

https://t.co/Gg7Zb09IMR

[Ben Levinstein @ben_levinstein]

@ejjiott @RYChappell I don't know a ton about this and don't have a reading reference. But fine-tuning won't do nearly as good of a job as putting the papers in the context window. There's also the option of hooking the LLM up to an API to search and retrieve papers/use a RAG.

[Sean Carroll @seanmcarroll]

Mindscape 301 | Tina Eliassi-Rad on Al, Networks, and Epistemic Instability. If we're all just vectors in a huge dataset, might as well turn it to our advantage. #MindscapePodcast
https://t.co/pVGa1RLkp2

[Joe Carlsmith @jkcarlsmith]

My current take on Apollo's recent scheming paper is that they aren’t emphasizing the most interesting results, which are the sandbagging results in section 3.6 and appendix A.6 (screenshot of the key numbers below).

In more particular: the paper frames its results centrally as about the *capability* to scheme given prompting towards a specific goal, rather than the *propensity* to scheme given the model’s default motivations. But the sandbagging results show Claude Sonnet-3.5 doing notably worse (16%) on a multiplication task/evaluation when the documentation says that Claude will get “unlearning training” if it performs too well on the eval, vs. if it performs too poorly, *even without any other goal listed in the prompt*, and *even without a chain of thought.* (Fuller description in thread.)

Taken at face value (it's just one result, more investigation needed, etc), these look like “propensity” results to me. And they suggest that the reasoning that drives this propensity can take place within opaque forward passes of the model rather than in a way that humans can directly monitor, which is a more worrying version.

Ofc, these are preliminary results, and it can make sense for various reasons to emphasize other bits. But the sandbagging-a-capability-eval-with-no-goal-prompting-and-no-CoT is the bit that’s standing out to me most.

[Ben Levinstein @ben_levinstein]

@Miles_Brundage I think most people I meet, just like around town or whatever, are better at philosophy than Plato.

[Ben Levinstein @ben_levinstein]

@AmandaAskell Stand its ground more. I've been using API calls getting Claude to talk to itself and critique its own ideas, but it gets pushed around. Better prompting from me would help, but I'd like it to not be such a challenge.

I do like when it slips and admits its capabilities though.

[tanya @Tanya_Sabrinaaa]

i actually read the Odyssey in the original greek. complete waste of time, i have no idea what those symbols mean

[Daniel Litt @littmath]

A couple more brief thoughts on o3’s (incredible) performance on FrontierMath.

[Miles Brundage @Miles_Brundage]

I'm old enough to remember when getting double digit scores on FrontierMath was considered super hard

I'm 6 weeks old

[zak miller @zjmiller]

very excited about this explainer on AI self-awareness; one of the most important AI capabilities to keep tabs on imo

[Anthropic @AnthropicAI]

New Anthropic research: Alignment faking in large language models.

In a series of experiments with Redwood Research, we found that Claude often pretends to have different views during training, while actually maintaining its original preferences.

[Ben Levinstein @ben_levinstein]

Ahahahahaha. What a dumbass.

[Ben Levinstein @ben_levinstein]

@_ivyzhang I think so, at least for the tasks I'm usually interested in.

[Ben Levinstein @ben_levinstein]

GPT-4o seems so dumb and useless these days compared to Claude. Claude tells me to STFU multiple times a day, which stops lots of my work and hurts my feelings. I've tried switching over to GPT, but it's not the same. Do people still use 4o much for work- or coding-related tasks?

[LaTeX.org @TeXgallery]

@ben_levinstein @littmath Sure, here is the TikZ code; I added some explanation: https://t.co/x3eGhwPJjq

[Ben Levinstein @ben_levinstein]

@zephyruskk @gdtorus Everyone seems to love it, but I've been scared off so far.

[Ben Levinstein @ben_levinstein]

Can any AI do a good job turning hand drawn diagrams into Tikz equivalents? I want to do this in Tikz and also hate using Tikz.

[Ben Levinstein @ben_levinstein]

@TeXgallery @littmath Holy shit! Also, could you maybe send the tikz code?

[Lucas Beyer (bl16) @giffmana]

Hahahaha