@zutalorz @clarissawam @stevo887 Exactly - except the classes of problems that ML is used for aren't necessarily checkable in that way. Take pharmaceutical design for example - Thalidomide was a tragedy, but caught and understood because there was an understanding of the underlying biochemistry, so they spotted to problem. But when you've been handed a flawed "solution" without doing the underlying research it may be orders of magnitude harder to find the failure point.
=> More informations about this toot | More toots from _thegeoff@mastodon.social
text/geminiThis content has been proxied by September (3851b).