Lightnews — Scholar-powered news

Reposted by Claus Wilke

Claus Wilke @clauswilke.com · 21h

Biology is much more complicated than most non-biologists can imagine. And AI is not going to change this anytime soon.
blog.genesmindsmachines.com/p/we-still-c...

We still can’t predict much of anything in biology

Biology is hard. Yes, even for AI.

blog.genesmindsmachines.com

4 40 110

Claus Wilke @clauswilke.com · 53m

In the mean time, a good proxy is how quickly it starts producing designs. If you've got to wait forever and nothing is happening you may have to switch the target. If you're getting results quickly you're probably good.

Claus Wilke @clauswilke.com · 54m

It's too early to make statements about which binding pockets work better or worse. We see there are differences but haven't quantified them in any way. We'll probably write this up as a paper eventually.

1

Reposted by Claus Wilke

Titus Brown @titus.idyll.org · 19h

Really good post!

Claus Wilke @clauswilke.com · 21h

Biology is much more complicated than most non-biologists can imagine. And AI is not going to change this anytime soon.
blog.genesmindsmachines.com/p/we-still-c...

We still can’t predict much of anything in biology

Biology is hard. Yes, even for AI.

blog.genesmindsmachines.com

1 1 11

Claus Wilke @clauswilke.com · 21h

@harmitmalik.bsky.social

Claus Wilke @clauswilke.com · 21h

I'd say even with available training knowledge success is not guaranteed. See the wide spread in performance in supervised prediction of mutational effects.

1

Reposted by Claus Wilke

Gaurav Sablok - I will immediately accept submitted job offer @omicscode.bsky.social · 21h

Thats very true. I even posted a post on my X account about that. People thought that AI is there so we don’t need interpretation but they didn’t realized that AI is learnt from the available knowledge and not that something which is not present.

Claus Wilke @clauswilke.com · 21h

Biology is much more complicated than most non-biologists can imagine. And AI is not going to change this anytime soon.
blog.genesmindsmachines.com/p/we-still-c...

We still can’t predict much of anything in biology

Biology is hard. Yes, even for AI.

blog.genesmindsmachines.com

1 1 3

Claus Wilke @clauswilke.com · 21h

It's quite possible that some of the binders are binding and just not giving us a signal in our assays. We're aware of this possibility and thinking about ways to test it.

Disclaimer: This is a large-scale effort with multiple labs involved. My role is computational.

2

Claus Wilke @clauswilke.com · 21h

Biology is much more complicated than most non-biologists can imagine. And AI is not going to change this anytime soon.
blog.genesmindsmachines.com/p/we-still-c...

We still can’t predict much of anything in biology

Biology is hard. Yes, even for AI.

blog.genesmindsmachines.com

4 40 110

Claus Wilke @clauswilke.com · 1d

I think people have to be careful with target-date funds. Being 70% in cash and bonds by year 65 is a terrible idea if you think you'll live to your late 80s or early 90s. It's probably better to just buy SP500 or Total World Stock Index or similar.

1

Reposted by Claus Wilke

Shabaana Khader @khadershabaana.bsky.social · 2d

Dear #microbiology and #immunology community- I took a break from social media after I deleted my Twitter account, but am back here. Please help me rebuild my community by following me and amplifying this message- I will follow you back. Thank you and I look forward to our many interactions!

5 32 71

Reposted by Claus Wilke

Frames @oxinabox.bsky.social · 3d

as a girl with a PhD in natural language processing and machine learning it's actually offensive to me when you say "we don't know how LLMs work so they might be conscious"

I didn't spend 10 years in mines of academia to be told ignorance is morally equal knowledge.

We know exactly how LLMs work.

65 1.6K 5.7K

Reposted by Claus Wilke

Erin Green @erinrgreen.bsky.social · 3d

Job alert ‼️ UChicago Micro is hiring! Open to tenured/tenure track faculty at all levels in any area of microbiology. Come join our amazing and growing department. apply.interfolio.com/174404

4 120 120

Claus Wilke @clauswilke.com · 5d

Some thoughts about the research-plan component of the NSF GRFP application.
blog.genesmindsmachines.com/p/how-to-wri...

How to write an NSF GRFP research plan

This is my second post about the NSF GRFP.

blog.genesmindsmachines.com

1 2

Claus Wilke @clauswilke.com · 6d

Yup, been like that for ages. I find it confusing because not only does the model change but the formula changes as well. And the formula for the GAM is not particularly intuitive if you're new to that type of model.

1 2

Reposted by Claus Wilke

Claus Wilke @clauswilke.com · 7d

In her latest video, @acollierastro.bsky.social tackles physics grifters and I'm so here for it.
www.youtube.com/watch?v=miJb...

conspiracy physics and you (and also me)

YouTube video by Angela Collier

www.youtube.com

4 13

Claus Wilke @clauswilke.com · 7d

In her latest video, @acollierastro.bsky.social tackles physics grifters and I'm so here for it.
www.youtube.com/watch?v=miJb...

conspiracy physics and you (and also me)

YouTube video by Angela Collier

www.youtube.com

4 13

Claus Wilke @clauswilke.com · 7d

In retrospect I should not have used the MPI mechanism. I thought I needed to go that route because I needed an experimental collaborator, but these days I would just go in as sole PI and make any collaborators senior personnel.

1 1

Reposted by Claus Wilke

Craig Kaplan @triggerloop.bsky.social · 7d

I think the R01 is basically a non-starter at NIGMS now. Data from NIH Reporter for new and competing awards, comparing number of R01s issued relative to R35s (MIRA). This has major ramifications for basic science and basic science investigators that are not yet in the MIRA pool

Graph showing total R35+R01 awards issued by NIGMS from 2020 to 2025. R01 in black showing a steep decline. R35 showing an increase. Total awards in 2025 slightly up from 2024 but these years are lower than 2020-2023

11 43 72

Claus Wilke @clauswilke.com · 7d

Same here (zero path towards MIRA). I couldn't apply when I had an NIGMS R01 because it was an MPI grant, and now it's run out and I can't apply anymore either. I was funded by NIGMS for almost 20 years but I assume it's over now.

1 2

Claus Wilke @clauswilke.com · 8d

I have seen numerous successful and not so successful applications. I can't share them because they're not mine, but I've written a blogpost that may be helpful. I may write a second one later this week, about the research plan.
blog.genesmindsmachines.com/p/how-to-wri...