Content
@
0 reply
0 recast
0 reaction
July
@july
a crazy idea: the fact that most (if not close to all) models, papers, are all going to be obsolete in a few years even though they feel so new today
7 replies
2 recasts
22 reactions
Claus Wilke
@clauswilke
I mean, if you've been in the space for a while (20+ years for me) you've seen it all over and over. I remember in the late 90s, wavelets where the thing for image processing. And now most people probably have never even heard that term.
1 reply
0 recast
0 reaction
Ben 🟪
@benersing
What advanced research has you most excited in computational biology?
1 reply
0 recast
0 reaction
Claus Wilke
@clauswilke
I'm very excited about the application of large language models to problems of biochemistry. There are so many novel applications. For example: https://www.biorxiv.org/content/10.1101/2023.11.15.567263v1.abstract
1 reply
0 recast
0 reaction
Claus Wilke
@clauswilke
And also this one. This one is particularly nice because Chamath on the All-In podcast said (a few months ago) that if a machine learning model could ever do something like this it'd be the downfall of civilization. 😉 https://www.sciencedirect.com/science/article/pii/S2666389923002490
1 reply
0 recast
0 reaction
Ben 🟪
@benersing
Interesting. Sounds like I have to read it!
1 reply
0 recast
0 reaction
Claus Wilke
@clauswilke
The reality is always more boring though. We just found that with clever prompt engineering we can recognize molecules with similar functions that don't have immediate obvious similarity in chemical structure.
1 reply
0 recast
0 reaction
Claus Wilke
@clauswilke
The key idea is that there are alternative, equivalent string representations of the same chemical structure and chemical large language models can somehow see underlying chemical concepts when provided with these alternative representations.
1 reply
0 recast
0 reaction
Ben 🟪
@benersing
What's been your biggest learning in your prompt engineering journey?
1 reply
0 recast
0 reaction
