Raw LLM Responses

The idea that AI will solve cancer or whatever is nonsense. AI works well for coding and math because they are fundamentally deductive: I can formulate a problem and prove that the solution I arrived at is correct. Biology does not work like that. Biology is the accumulation of billions of years of random mutation and selection, and there is not underlying optimization function that can be inferred (species don't evolve towards "better" versions of themselves, whatever that may mean). Biological processes still obey physical rules, and one could imagine distilling human biology into an inconceivably complex network of chemical reactions subject to stochastic noise (something that, while daunting, would still be in principle simulatable by a sufficiently deep and wide neural network). But someone has to go in an measure all the data that would be needed to describe it. As Mr Bores is aware, this would require not only that scientists reorient themselves to work 24/7 on gathering these data, but to also agree on universal measurement standards and to commit to publishing all of their results on a database that was accessible to those who might want to build these models. I don't know how one does any of that without massive government funding, and with stringent regulations that protected the medical information of all the test subjects involved. As lowish hanging fruit we could start by digging into the data collected by all the major pharmaceutical corporations. But that is heavily guarded, proprietary information, that is most likely incomplete, with databases that are not compatible across companies. Not to mention all the data that has been lost every time some startup tried something and failed, or every time someone didn't write down something that didn't seem pertinent at the time. The scientific literature is littered with absolute garbage, and we'd need to devote an enormous amount of effort to validate and replicate the published record if we are to feed that into any kind of predictive model. Unfortunately for us, capitalism has made most of our jobs quantifiable, which is all ML needs to replicate them. So we will all be out of a job well before these companies make much of a dent on human health.