+1. This paragraph in particular is such a helpful frame, a yardstick for measuring the actual importance of new capabilities and models as they arise – something which hype otherwise makes difficult.
> If the model tells you to do something you would probably have done anyway, it’s useless. If the model replaces something you would have needed to do manually, it’s somewhat useful. If the model increases your odds of a successful therapy, it’s extremely useful, and if it adds successful therapies it’s world-changing.
Thanks for a great post. Given the speed of progress at various levels in the AI stack, I wonder if or when we'll progress from these point solutions to more of a comprehensive simulation of what happens in a living organism. It seems like we're still at the stage of looking under the street light, but I bet we'll be able to fill in the gaps at an accelerating pace.
Nice, comprehensive overview of available AI models, thank you. It had led me to an idea for my own potential research.
Speaking of your AI conclusions, I agree with them in short to medium timescale. However, I am expecting that we will be able to model the whole organism and design a drug that we can give to patients right away. But when I say this, I think about what we will be capable in the end of this century or a few centuries. If we are talking about closest years and 1-2 decades, you are right to say that there are too many unknown-unknowns.
And I'd like to add that apart from direct modeling of biology, I expect great advancements in rate of discovery in science, including drugs, to happen because AI will help us browse all of our scientific knowledge and humanoid robots will help conduct experimentation faster and with lower error rate. These two factors that are derived from AI will increase researcher's productivity. By increasing rate of scientific discovery, this will also facilitate greater opportunities for therapies development
Some interesting work involving AI is underway in a $68 million neuropsychiatric research award round financed by California's stem cell and gene therapy agency. Grants were awarded last week. Read about it on the California Stem Cell Report.
This is very interesting. How effective is AI in automating experiments to be cheaper and faster? Can you just get to some sort of highly efficient combinatorial chemistry factory that just churns out drugs for arbitrary diseases?
Thanks so much for this, Sarah! I was aware of small, piecemeal components of these such as ESM and AlphaFoldx, but greatly appreciate this survey of the overall current AI-bio landscape!
Great read. The analytical frame set up at the top of the piece is informative and clarifying. Thanks!
+1. This paragraph in particular is such a helpful frame, a yardstick for measuring the actual importance of new capabilities and models as they arise – something which hype otherwise makes difficult.
> If the model tells you to do something you would probably have done anyway, it’s useless. If the model replaces something you would have needed to do manually, it’s somewhat useful. If the model increases your odds of a successful therapy, it’s extremely useful, and if it adds successful therapies it’s world-changing.
Thanks for a great post. Given the speed of progress at various levels in the AI stack, I wonder if or when we'll progress from these point solutions to more of a comprehensive simulation of what happens in a living organism. It seems like we're still at the stage of looking under the street light, but I bet we'll be able to fill in the gaps at an accelerating pace.
Nice, comprehensive overview of available AI models, thank you. It had led me to an idea for my own potential research.
Speaking of your AI conclusions, I agree with them in short to medium timescale. However, I am expecting that we will be able to model the whole organism and design a drug that we can give to patients right away. But when I say this, I think about what we will be capable in the end of this century or a few centuries. If we are talking about closest years and 1-2 decades, you are right to say that there are too many unknown-unknowns.
And I'd like to add that apart from direct modeling of biology, I expect great advancements in rate of discovery in science, including drugs, to happen because AI will help us browse all of our scientific knowledge and humanoid robots will help conduct experimentation faster and with lower error rate. These two factors that are derived from AI will increase researcher's productivity. By increasing rate of scientific discovery, this will also facilitate greater opportunities for therapies development
Some interesting work involving AI is underway in a $68 million neuropsychiatric research award round financed by California's stem cell and gene therapy agency. Grants were awarded last week. Read about it on the California Stem Cell Report.
https://david293.substack.com/
This is very interesting. How effective is AI in automating experiments to be cheaper and faster? Can you just get to some sort of highly efficient combinatorial chemistry factory that just churns out drugs for arbitrary diseases?
Thanks so much for this, Sarah! I was aware of small, piecemeal components of these such as ESM and AlphaFoldx, but greatly appreciate this survey of the overall current AI-bio landscape!