In honor of the 50th anniversary of the Apollo moon landing, it's a pleasure to share the great accomplishments of the Atomwise team! #drugdiscovery#ai#moonshotlnkd.in/gwtP3Qk
While waiting for the grandkids to show up for Christmas dinner later on today I decided to work on my science fiction roleplaying game.
I typed almost a whole paragraph of gibberish. One sentence read:
Yjod eo;; nr s vjs;;rmhr brtdid yjr rmr,ud Omyr;;ohrmvr
My fingers were one step off on the keyboard, and now I have no way to recover what I wrote in my inspiration except to paintakingly move my fingers one step over and try to re-type the gibberish. So I guess Happy "Yjod eo;;" to you all, and to all a good night.
Be the optimizer. Train MNIST—by hand.
Ride a saddle into a valley, watch FP16→FP32: crunchy → smooth.
Zoom step size and see curvature, tweak batch size, or run SGD if you can’t descend by hand.
Live (desktop is better) @ humandescent.net
Have you ever done a dense grid search over neural network hyperparameters? Like a *really dense* grid search? It looks like this (!!). Blueish colors correspond to hyperparameters for which training converges, redish colors to hyperparameters for which training diverges.
Scoop: Atomwise has hired a CMO w/ 25-plus years experience in the drug industry, as the AI biotech pivots from partnerships to pipeline
CEO Abe Heifets talked with me about the new focus, starting with TYK2, and why faster isn't necessarily better
endpts.com/atomwise-ceo-o…
Only 3 days left until Ai4 2023 in Vegas on August 7-9. Our panel “AI Powered Pharmaceuticals: Intelligent Drug Discovery And Clinical Trials” will gather pharmaceutical leaders to discuss the ways AI is making an impact on Pharma. Passes are selling fast: ai4.io/usa/
My #Tecbio list.
This is my list of companies using #AI and #ML in biotech drug discovery.
1. $ABCL
2. $RLAY
3. $SDGR
4. $RXRX
5. $EXAI
6. $ABSI watch list
7. @AtomwiseInc private
If you have a name I am missing please let me know.
Your top pick in #Techbio is?
@KRHornberger If drug discovery is done in humans, then you're making a stronger claim, to wit: "Our understanding of disease biology is simply too poor to model clinical outcomes right now". If you're right about _that_, then the concern about AI is a simple corollary.
@AbeHeifets I’m not sure that it would. A wise pharma executive I know routinely says: drug discovery is done in humans. Our understanding of disease biology is simply too poor for AI to model clinical outcomes right now. Wouldn’t mind if that changed someday though.
@KRHornberger Put another way, and leaving AI out of it, do you think that adding (or subtracting) 50% to a preclinical budget in dollars or time would change the clinical failure rate?
@KRHornberger Some clinical trial failure rate is due to imperfect correlation between preclinical models and clinical outcomes, and so would remain even if we find the "best" molecules (according to preclinical models). But do you mean there's no chance to improve the quality of cmpds today?
“identity theft” is such a wild crime
It has nothing to do with someone stealing your “identity,” whatever that means, and everything to with companies that have poor access controls slandering you
and somehow society is programmed to be like, this makes sense