Bayesian Optimization is just slightly-smarter random search

I took both pills

I need to find more accounts on Bluesky in AI or Bio to follow. My 'Following' feed here is totally dead

This perspective on AI slop was pretty interesting. Their description of using current AI models to do creative tasks is roughly in line with my own experience.

"The choice of the problems to study", key question for progress in science.
#NobelLectures

Big fan of reading yearly reviews and especially enjoyed this one by
@ftlsid.com . Think I'm inspired to (for the first time) write my own.

cybermonk.substack.com/p/26

Had a productive evening planned but saw a new Talon Fitness nutrition tier list in your feed...

That's my evening gone I guess

I think taxation also falls into a similar category to climate change, where really solving it requires international cooperation. As long as there is one place willing to undercut everyone else's tax rates, tax rates everywhere will get pulled down. Been a problem in Europe at least for a long time

Can AI improve the current state of molecular simulation?

www.owlposting.com/p/can-ai-imp...

in my first podcast, I spend 2 hours interviewing Corin Wagen and Ari Wagen, two brothers who are building the next generation of molecular simulation for drug discovery and material science

📚💙 Just finished David Bessis' book 'Mathematica' and it was very thought provoking. Think I will definitely need to reread at some point to really absorb it. Talks about how math is misrepresented as this dry logical reasoning game while it is really a forest of imagination and creativity

This paragraph on talent in bio by Sam Rodriquez is interesting. My experience in ML is actually frustration at these 'high prestige' research paths in industry, because I see friends go and work as 'research scientists' there when I think they could have much more impact doing something else

Better ML techniques will indirectly result in improvements to hardware for performing biological experiments.

Better ML techniques -> we can pick more valuable experiments right now -> demand to do the experiments increases -> capitalism does it's thing and people invent faster, better hardware

Paragraph in the Amodei essay that I agree with. Seems like there is a very strong power law in Biology: a very small number of technologies that have totally shaped what kinds of questions we go after and how we answer them

30 mins into the essay and finally found something Dario doesn't think AI can do in '5-10 years': mind-uploading...

Learnt about the Liftware Level - a spoon that balances out tremors: www.youtube.com/watch?v=kQRv...

Not a mechanical engineer so was a bit eye opening to see that such a thing was even possible

Concerning: website returns a 404. Hopefully just a temporary outage and they didn't go bankrupt...

"Perhaps there is some way to coax the adult brain into an earlier or more plastic state where it can be reshaped. I’m very uncertain how possible this is, but my instinct is to be optimistic about what AI can invent here."

From Machine's of loving Grace by Dario Amodei

Came over to check out Bluesky. Will make some of my usual posts here for a while to see how I like it. Intro:

I'm a machine learning PhD student who is interested in how to use machine learning to accelerate scientific discovery

I recently started casually blogging at substack.com/@algobaker