Drug development currently involves the use of AI to search for new compounds but still relies on various phases of testing on human trials to ensure efficacy and safety.
An option to skip human testing would be to rely on accurate simulations of the human body, but this would require advanced models that can accurately represent the complex interactions within human biology, including genetic variations and long-term effects. Achieving this requires vast computational power, sophisticated algorithms, a deeper integration of diverse biological data and obviously overcoming regulations.
As far as I understand, the computational requirements for such an effort would be in the realm of exaflops, not far from current supercomputers.
What are the odds to have an AI generated drug that skips human testing within a decade?
Feel free to comment and share your thoughts and insights!
(The drug should be approved by a major health organization to resolve true.)
1,000
3.00I think this happens if there is a very convincing drug that targets one thing and one thing only, and there is a high degree of confidence that there are little to no off-target effects. It would also have to be relevant to a disease that many people are suffering from, in order to apply enough collective political pressure on state institutions. This probably makes this more likely to occur in a smaller nation state with advanced medical technology and weaker regulation, so perhaps somewhere like UAE or Israel? Countries that are big players in the medical tourism scene in the 2030s are also likely candidates, e.g. Turkey, Brazil, Thailand
@MalachiteEagle in the description I mentioned a "major" health organization. Although vague, I'd approve it if it supervises drugs in a region with over 200 m people, like the USA, china, the EU, India and Brazil
@SimoneRomeo Seems like a reasonable judgement. I just wanted to clarify since it seemed like that contributes a lot of likelihood relative to getting the computational biology fully worked