zlacker

Can you imagine spending decades of your life studying antibiotics, only to have an AI graph neural network beat you to the punch by conceiving an entire new class of antibiotics (first in 60 years) and then getting published in Nature.

https://www.nature.com/articles/d41586-023-03668-1

>>tansey+(OP)
It looks like the published paper managed to include plenty of citations.

https://dspace.mit.edu/handle/1721.1/153216

As it should be.

>>tansey+(OP)
As you already know yet are being intentionally daft about: They didn't use an LLM trained on copywritten material. There's a canyon of difference between leveraging AI as a tool, and AI leveraging you as a tool.

LLMs have, to my knowledge, made zero significant novel scientific discoveries. Much like crypto, they're a failure of technology to meaningfully move humanity forward; their only accomplishment is to parrot and remix information they've been trained on, which does have some interesting applications that have made Microsoft billions of dollars over the past 12 months, but let's drop the whole "they're going to save humanity and must be protected at any cost" charade. They're not AGI, and because no one has even a mote of dust of a clue as to what it will take to make AGI, its not remotely tenable to assert that they're even a stepping stone toward it.