[namaria]

Non-deterministic computer programming (which is conceptually what ML/AI is: brute force the mapping from input to output upfront) sounds great and can truly unlock whole areas of problem and solutions spaces we didn't even know existed. Until you realize you're just creating infrastructure where you cannot possibly hope to determine behavior. Something that is fundamental to debugging and security.

[blablabla123]

Also I wonder how much scientific rigour went into the recent progress in AI/ML. At least in principle there might be ways to determine invariants and boundary conditions (without modifying models), also considering there had been efforts to change ML models and make them explainable.

[namaria]

I think all of this will be rendered bunk with the erosion of the cheap compute that underlies this current boom in cloud/saas/ml/ai. It all rests on printing dollars and shipping them to China and will implode as soon as the CCP decides they need to militarize the Taiwan matter. Every cottage industry dependent on an exploding number of datacenters should fall in line with the rest of the economy then and we'll be back to resource constrained software.