@krishnanrohit this is their active and explicit goal which they are making measurable and sustained progress on. what part is ludicrous?

@tenobrus If they do pharma and build another Eli Lilly, great. But it's not going to be all by themselves, and will be at the expense of not building a giant cyber firm, or materials,or a consulting firm, or a rocket company, or fundamental physics, or million other areas that also exist.

@JacobWitten1 @krishnanrohit i meant this not as evidence that ai was currently capable of massively outcompeting existing bio companies, but rather as evidence that anthropic is in fact iterating on it for such purposes and is doing so with internal employees and perusing drug discovery on their own

@tenobrus @krishnanrohit FWIW this is extremely "normal technology" stuff akin to acceleration in biotech from next-generation sequencing, and note that Illumina is a niche player not world-bestriding colossus

@tenobrus @krishnanrohit Drug design is like 0.01% of bringing a new drug to market.

You can walk the halls of any university and find 500 drugs that have been designed by researchers. The value is close to zero.

@tenobrus @krishnanrohit Protein design has ground truth in a way most of biology doesn't so nobody doubts it'll get really good with AI

@krishnanrohit @tenobrus tbh paywalling biopharma capabilities and charging a lot for it sounds like a good enough way to capture a lot of value from the industry; dont necessarily need to verticalize

@maxisawesome538 @krishnanrohit i'm aware!

@tenobrus @krishnanrohit Mostly agree with this, though unknown at least to me what their wet lab scaling plans are and without scaling wet lab they're gonna need the existing ecosystem

@virtu101 @tenobrus Capture how. Who's going to use it just to provide them training data for free.

@tenobrus @krishnanrohit there is so so much more to the drug discover process than generating candidates. A huge majority of the cost comes later

@krishnanrohit @virtu101 for free? companies can just buy other companies

@tenobrus @virtu101 If the argument is that anthropic will buy a pharma co to build a better pharma co, yes it's possible. Even if all they end up doing is workflow optimisation using Claude. But it's not going to be the sole winner.

@krishnanrohit they are going to eat every single business in existence a failure to do so means a failure to achieve ASI

@tenobrus @krishnanrohit sure, im notextrapolating that far based on a few evals, but we shall see!

@tenobrus @krishnanrohit Of course this will all sound silly once they start tiling the Mojave with self-replicating biotech research labs but you know, for the next 2-3 months until the Mojave is fully tiled they will have to work with Lilly

@proales @krishnanrohit sure

@krishnanrohit this is one of those heuristics that almost always works

@krishnanrohit You must Believe in the country of geniuses. Millions of machines humming. Biology, physics, history…

Just throw the computer at it.