JCs are drooling at the prospect I assume.
3 Likes
AI is best as sorting though existing data and spotting relationships that suggest new paths.
To me PhD breakthrough suggests creating new data. They are using AI to suggest drug candidates that fit better with folded enzymes. New wonder drugs could result. But we have been though half a dozen rounds of this kind of wonder for new drug development. Only a few succeed.
Ask AI to suggest a new wonder alloy for steel. What are the odds? Is there any combination of metals that has not been tried? Did they somehow overlook its wonder properties last time? Not very likely.
I think we will still require well trained professionals to sort through the AI suggestions and decide which are worth trying. That means fewer professionals but you still need to train replacements and requirements are more demanding.
1 Like
If you follow the Stuart Kauffman lectures, my favorite Complexity scientist, you realize that new developments build on old ones. It could be accidents…
In 1839 he (Goodyear) accidentally dropped some India rubber mixed with sulfur on a hot stove and so discovered vulcanization.
or, in evolution, it’s the Adjacent Possible
There is nothing new under the Sun, just variations on old themes.
Another way of looking at it is to think of random events a few of which, very few of which, manage to survive, at least for a time. Edison is another good example, how many times did he fail before getting a light bulb to work? How often have SpaceX rockets blown up yet they brought back stranded astronauts.
The Captain
2 Likes
Ok, but the Edisonian approach to research is try everything on the shelf. Insight into which to try first. Or design of the ideal solution can be much more efficient.
Chemical industry has a history of screening programs. Try everything in my library of materials to see if any perform in an area of interest. Then test another library. This is the way Roundup was discovered. It gives the best patent protection.
Patent office often finds that invention expected to work is obvious to one skilled in the art. Weak patent or very limited coverage results. AI needs deep thinking and unique approach to succeed.