Tim Hua 🇺🇦

AI safety, Econ, new liberalism, math, and a bit of art history (as a treat)

Behavioral evaluations @TransluceAI. Prev Astra, MATS & Walmart's Econ Team

[Neo Research @NeoResearchAI]

We're Neo Research (新衡). Asia’s first independent frontier AI safety evaluation & research lab.

Today we're publishing our first report: an independent safety evaluation of DeepSeek v4 Pro. (1/5)

[Tim Hua 🇺🇦 @Tim_Hua_]

From this lawsuit

(Although a true LW enjoyer would note that ChatGPT is definitely built with safety “in mind,” which is a much lower bar than being “safe.”)

https://t.co/fwLdeCrfJB

[TechCrunch @TechCrunch]

Florida sues OpenAI, Sam Altman, in first-of-its-kind lawsuit over violent incidents https://t.co/VzYdYraxpN

[Tim Hua 🇺🇦 @Tim_Hua_]

Lesswrong coded footnote

[Tim Hua 🇺🇦 @Tim_Hua_]

We also see this in humans (my younger brother never remembers what I tell him to do no matter how many times I say it.)

[Christopher Potts @ChrisGPotts]

Remarkably, we see the same gradient interference pattern we saw with the idealized tasks. For example, here we inject task examples every 100 batches. For the largest models, the task loss drops at these injection points. The task is then partially overwritten (loss goes back up), but the overall loss trends downward, which corresponds to successful learning. The small models never get traction.

[Tim Hua 🇺🇦 @Tim_Hua_]

@wilhelmscreamin I’ve never read planecrash 🫣

I feel like I really don’t need another one of those in my life rn

[Tim Hua 🇺🇦 @Tim_Hua_]

Ur ngmi unless you know your way around the extended hpmor universe

[dave kasten @David_Kasten]

@TheStalwart @AgustinLebron3 @lion_tender I regret to inform you that it is, actually, a fairly important source if you want to understand the history of the AI companies and their ideology.

[Tim Hua 🇺🇦 @Tim_Hua_]

@daniel_271828 @BogdanIonutCir2 Ah yeah I guess evidentially it’s good. Causally it’s not clear.

[Tim Hua 🇺🇦 @Tim_Hua_]

I think it's good that OpenAI felt the need to distance itself from LTF in public.

[OpenAI Newsroom @OpenAINewsroom]

OpenAI’s approach to AI policy and political advocacy, including how we represent our policy views publicly and why we believe AI policy debates should be transparent

https://t.co/MKsTWfCWLH

[Tim Hua 🇺🇦 @Tim_Hua_]

@luke_drago_ Perhaps we should introduce the broader public to 0DTE options trading

[Tim Hua 🇺🇦 @Tim_Hua_]

@EconTheory @ecmaEditors @econometricsoc Please change it back

[Tim Hua 🇺🇦 @Tim_Hua_]

They neutered econtheory dot org

[Tim Hua 🇺🇦 @Tim_Hua_]

I've repeatedly said that work in AGI political economy feels much more neglected and important than AGI economic growth theory. At this point I'm just happy that anyone is working on it!!

[Matthew Botvinick @mattbotvinick]

Anthropic now has a team dedicated to AI and the rule of law — and we've just opened our first role.

@AnthropicAI has studied what AI means for the economy. This team asks a different question: what will it mean for executive power, for courts and elections — and for the public deliberation that constitutional democracy ultimately rests on?

We're looking for someone with real depth in both AI and the law — a legal scholar, political scientist, or experienced government hand who can reason about frontier systems and the institutions they will affect.

If that's you, or someone you know: https://t.co/668HDz1lhf

[Tim Hua 🇺🇦 @Tim_Hua_]

From: https://t.co/w2YZGinXFR

See also @AndersCWoodruff: https://t.co/DLWTRL2yay

[Tim Hua 🇺🇦 @Tim_Hua_]

When people talk about the cost of running AI systems, I often refer them to this fact: anything an AI could do today, it can do in a year for 1/20th of the price!

[Tim Hua 🇺🇦 @Tim_Hua_]

@kevinroose @TheStalwart @lion_tender Kevin have you read it

[👣ℙ𝕖𝕕𝕣𝕠'𝕤 𝕄𝕦𝕤𝕥𝕒𝕔𝕙𝕖🇺🇲 @OfAthenry]

Crazy things are happening in the drone world

[Tim Hua 🇺🇦 @Tim_Hua_]

@testingham Yeah maybe? I mean there are also just much more frictions to scaling labor.

The broader point (which I think you agree with) is that we can see what happened with Mythos & cyber more broadly across the economy soon.

[Tim Hua 🇺🇦 @Tim_Hua_]

This is an example of how increased capital productivity leading to higher capital share.

Before Mythos, you probably *couldnt* spend a million dollars to beef up your security even if you wanted to! But now you can.

[Peter Wildeford🇺🇸🚀 @peterwildeford]

Mythos at Palo Alto Networks "found more than two dozen critical vulnerabilities in around three weeks, roughly five times what the company would typically find using existing tools"

But the company "burned through more than $1 million worth of tokens using Mythos"

[Tim Hua 🇺🇦 @Tim_Hua_]

@eccentric1ty I think having those would bring back fond memories of being back in China (where these things are somewhat common) for me, and the fact that they have some chance of stopping killer drones makes me like them even more.

[Tim Hua 🇺🇦 @Tim_Hua_]

oh we need this at lighthaven before the singularity

[Marshall Richards @marshallrichrds]

this was an intrusive thought that deserved to be tested.

drone vs. wood beaded door curtain

low-cost indoor drone defense has been achieved internally.

[roon @tszzl]

the frontier labs don’t have “comms problems”. reality right now has a comms problem. what is happening is a little scary and there’s no nice words anyone could say, especially not those profiting from it, that’ll make it feel that much better