[limflick]

I guess an optimistic way to look at this would be to treat this as just another layer of abstraction, meaning people could focus on larger scale problems moving forward, similar to how the evolution of programming languages influenced development time, quality and the quantity of software being put out. The question is at what price does all of this abstraction come at, assuming AI continues to evolve at its current rate.

[master-lincoln]

This can not be seen as layer of abstraction as it's non deterministic and not trustworthy. So we still need to inspect and understand that abstraction layer output if we want to have a reliable product

[bayindirh]

Adding non deterministic layers on top of a painfully deterministic layer to make more betterer deterministic things is an oxymoron.

...and many people choose to ignore that fact.