Anqi Qu

rhodes scholar, phd in ml, ex-quant
ฅ^•ﻌ•^ฅ
just trying to understand ai before ai understands me

[Anqi Qu @Anqinator]

@gpusteve 1x free will

[Anqi Qu @Anqinator]

@ishanskhare Letsgooooo Ishan!! Very cool work 🙌

[Anqi Qu @Anqinator]

@mandylu The chance is low, but never zero… stay safe 👀😭

[Anqi Qu @Anqinator]

@giovanni_ 🤡😭

[Anqi Qu @Anqinator]

World models for biology!!! Extremely cool work from some extremely cool people 🧬 proud of my friend @ayushnoori

[Ayush Noori @ayushnoori]

Simulating the consequences of an experiment before running it is a holy grail of AI for biology. How can we get there?

Our new paper, “World models for biomedicine,” which I led together with @njwfish, is out now in @CellCellPress’s special issue on AI in biology. We explore how biomedical world models could help scientists navigate the enormous space of possible biological interventions. In principle, they could:

🧪 Simulate how biological systems respond to interventions never previously tested

🎯 Search for interventions that drive cells, tissues, or patients toward desired states

🧭 Plan experimental campaigns over long time horizons

🤖 Power autonomous, closed-loop AI scientists

However, realizing this vision requires much more than calling a model a “world model.” What exactly defines a biomedical world model, and what capabilities distinguish it from related models? Where do today’s models already show promise, and where do they fall short? What new data and architectures are needed? What evaluations would demonstrate that a world model works?

We tackle these and other questions in this latest paper from @marinkazitnik's lab: https://t.co/musHiEcHMp

More below 🧵👇🏽 1/8

[Anqi Qu @Anqinator]

@mckinleylettre Don’t sleep on queens

[Anqi Qu @Anqinator]

In case anyone was wondering what my thoughts are on AI taking my job (kinda)

https://t.co/mL0LCLQRQf

[Anqi Qu @Anqinator]

Not working on higher math but the way I think about it is that for AI to solve open math problems, it needs qualities like the ability to think “creatively,” long-horizon logical reasoning, and have strong reliability guarantees. The main question boils down to is whether these capabilities translate directly to other domains like AI for cancer (which I think they do!)

[tender @tenderizzation]

my son max, short for softmax

[Anqi Qu @Anqinator]

@lihanc02 Omg excited to see what you’ve been working on!

[julia @juumelo_]

imagine having an arch nemesis and you’re only a regular nemesis to them

[Anqi Qu @Anqinator]

@ayushnoori @GRESEARCHjobs @UniofOxford This is incredible Ayush!! So well-deserved 🙌

[Anqi Qu @Anqinator]

@theombl Some to add to the banks list: Delta with their Skymiles. Also Roblox with their Robux

[Anqi Qu @Anqinator]

@bhat_suhaas I wonder if this mathematical “intuition” is partly just search efficiency aka knowing which paths are worth exploring without having to exhaustively formalise them

[Anqi Qu @Anqinator]

@cloneismin Yes! And it only gets harder to do so as AI is trained to be more accurate and benevolent

[Anqi Qu @Anqinator]

Last week I read MatrAIx which simulated 8.3 billion AI personas. It sent me back to a question I have spent a lot of time thinking about:

What would actually make a population of simulated humans scientifically interesting? https://t.co/caLQ768ujn

[Anqi Qu @Anqinator]

@giovanni_ So….. RL/eval startup when 🤩🤩🤩

[Anqi Qu @Anqinator]

@ayushnoori Working on it 🫡

[Anqi Qu @Anqinator]

Hi guys I made a new personal site over the past few days in the time I spent waiting for slowpoke Claude to finish its tasks. Check it out and let me know how many easter eggs you can find :)

https://t.co/ljUjxeWSwA