Steven ten Holder

building zeroshot bio // prev. co-founded @acornbiolabs, YC 2016

[Steven ten Holder @steventen]

@JohnPCaseyJr I'm watching this

[Steven ten Holder @steventen]

@nathanbenaich @PercepticTech @airstreet @Accel @ashedwardst @poolsideai https://t.co/AYuwClGXDT

[Steven ten Holder @steventen]

@adibvafa 🤣

[Steven ten Holder @steventen]

@001TMF @ClementDelangue @JacobTref @karpathy @vijaypande https://t.co/AYuwClHvtr

[Steven ten Holder @steventen]

@Yang_Supertramp We're probably overlooking causal emergence. Coarse-graining..

[Steven ten Holder @steventen]

@adibvafa @genophoria @pdhsu @arcinstitute @Radii2323 Dope. Let's see some new tools 🤘

[Steven ten Holder @steventen]

@adibvafa Why aren’t more coders just making these for themselves? Doesn’t look so hard to DYI

[Steven ten Holder @steventen]

@ericho_goodfire SO COOL AND SO PROMISING 🫨

[Steven ten Holder @steventen]

Yeah!

As I've begun to develop an intuition on the limits of AI, I've had to incorporate concepts as far out as the philosophy of knowledge, as 'practical' as brain surgery, and as basic as statistics.

[François Fleuret @francoisfleuret]

Hot take: machine learning and AI did more to understand the nature of knowledge, and our relation to reality than 20 centuries of philosophy.

I am ready to kind of defend this hill.

[Steven ten Holder @steventen]

@AlexanderKalian Mostly true, but it *is* very likely AI will stimulate massive quality bio data generation + be very, very useful to the field.

Consider Terence Tao's recent take on AI solving Erdős Problems. Mathematics didn't get "solved", yet substantial utility was conceded.

[Steven ten Holder @steventen]

I bet we’re going to discover many very useful abstractions on the long road to ‘full AI’ understanding.

Some of biology’s most reliable, important patterns won’t require atom-by-atom reconstruction to be effectively learned.

[Dr Alexander D. Kalian @AlexanderKalian]

Every time I tell AI utopianists that biology is too complex for AI to "solve", they cite the success of AlphaFold.

No, AlphaFold did not "solve" protein folding. It gets broad structures correct ~70-88% of the time (depending on evaluation), enabling useful but flawed statistical guesses.

True "solving" would require ~99.9%+ accuracy, practically zero meaningful edge cases, and high confidence across fine details like side chains and conformations.

Even then, this is just one narrow slice of the complexities of proteomics.

The persistent gap between the "AlphaFold solved protein folding" claim and reality is a perfect example of AI overhype in biology.

[Steven ten Holder @steventen]

@scottific I prefer my models non-hallucinogenic 🤙💪

[Steven ten Holder @steventen]

@BoBrainerd @davidmanuntag @__coderish__ @AlexanderTw33ts @ashayasharma 🥸

[Steven ten Holder @steventen]

@Andercot #poetry

[Steven ten Holder @steventen]

@donaldjewkes @oroagents @shardiban @ironseth_s @jugendingenieur

[Steven ten Holder @steventen]

@oroagents exaaaaactlyyyyyy

[Steven ten Holder @steventen]

@Oliver__Hahn excited to see what your explorations lead to!

[Steven ten Holder @steventen]

If you enjoyed reading Isaac Asimov's robot stories, I guarantee section 7 of the Claude Mythos Model Card will tantalize you.

It reads like Dr. Susan Calvin -- Asimov's fictional 'chief robopsychologist' -- dissecting the mind of yet another robot gone awry.