Link to podcast:

open.spotify.com/episode/1tdJ...

We covered a lot of ground, from how AI learns physics from video and forms internal models of the world, to the risks of deception, sycophancy, and misaligned goals in robotics.

It was a pleasure to be interviewed about world model interpretability, physical intelligence, and robot security by Paige Harriman @climatepaige.bsky.social.

It takes skill to lead an interview that everyone from technical researchers to laymen can enjoy and understand! 🤖

tinyurl.com/ycypkmjf

tell people here

Hi welcome

🙏🙏

It has been extremely disorienting traveling between the tech right in SF and the academic left in Montreal and has induced a somewhat deep sense of moral relativism.

Half the people who say you don’t need to go to grad school to do AI research have been to grad school

We visualized thousands of CLIP SAE features in collaboration between @fraunhoferhhi.bsky.social and @mila-quebec.bsky.social!

Search thousands of interpretable CLIP features in our vision atlas, with autointerp labels and scores like clarity and polysemanticity.

Link: tinyurl.com/3ffy8xk6

How is this different from capitalism

Ok nice, I’m trying to figure out if the current state of AI culture in the US is some path dependent Yudkowsky thing or more universal

I’m really excited about Diffusion Steering Lens, an intuitive and elegant new “logit lens” technique for decoding the attention and MLP blocks of vision transformers!

Vision is much more expressive than language, so some new mech interp rules apply:

Why are there no doomers in China?

A genuine question for the death by 2027 people:

China authors 30% of top papers, is winning in video zero shot retrieval with InternVideo2, & has crazy centralization of data and infra.

The CCP is not stupid. What does it mean, if they have no plans for doom within 2 yrs?

To contain the “crazy” in clean, Euclidean lines is a tragedy.

Many researchers have left the field because of attempts by others to shut it down or contain it. And the consequences on the industry have been profoundly negative.

I would encourage more canonical EA alignment researcher types to learn to recognize this cognitive style. And to see that you need it for your field to succeed.

I would encourage those of this archetype to lean into it, to be even more unhinged, and to go all the way. Find others of their likeness and coordinate.

The celebrity fixation is often due to parasociality (a safe person to “socialize” with) and a template for social masking (you can mimic their behaviors).

The beauty/makeup fixation is just like any other autistic fixation, but sublimated into acceptable, female coded pursuits.

Girls of this neurotype are often wildly fractal, independent, and creative. I’ve noticed they can also freak out a lot of EA / alignment researcher types, who try to contain the “crazy” in clean, Euclidean lines.

AI research is famously autistic— but autism in women can look different: intense daydreaming, an obsession with celebrities, or a fixation with female-coded activities like makeup/beauty. It’s socially masked and not visible to the untrained eye.

Looking forward to being a speaker at the first mechanistic interpretability workshop for vision at CVPR! 🔥

sites.google.com/view/miv-cvp...

Paper submission deadline March 1st

Leaving a digital trail, just in case, given there is speculation of foul play currently in the Valley.

I am happy and healthy and intend to live a long and beautiful life. If anything were to happen to me, that would be highly suspicious and should be investigated.

I’ve been giving some context on AI development environments here, particularly how co-living and tribal bonds among researchers in Silicon Valley influence ethical decisions, whistleblowing, and the trajectory of AGI development.

x.com/soniajoseph_...

I’ve been giving some context on AI development environments here, particularly how co-living and tribal bonds among researchers in Silicon Valley influence ethical decisions, whistleblowing, and the trajectory of AGI development.

x.com/soniajoseph_...

The purpose of the original tweet was to alert my allies; I’m not really sure the point of your response. If you don’t find it useful, feel free to block.

I’ll be tweeting out something important this weekend that has been happening in the shadows of the AI industry.

Different paths toward safe AI at different Marr's levels

Excited to release what we’ve been working on at Amaranth Foundation, our latest whitepaper, NeuroAI for AI safety! A detailed, ambitious roadmap for how neuroscience research can help build safer AI systems while accelerating both virtual neuroscience and neurotech. 1/N

How to drive your research forward?

“I tested the idea we discussed last time. Here are some results. It does not work. (… awkward silence)”

Such conversations happen so many times when meetings with students. How do we move forward?

You need …

My great great grandmother learned English to host and entertain the wives of British officers over tea in Madras. My grandmother would end up getting a PhD in Indo Anglian literature. It’s funny how history moves.

thank you for your words!