One of the first things I used ChatGPT (3.5 then) was to give it a JavaScrip algorithm I was writing and ask it to find the bug.

The answer surprised me. It started by saying something like ‘you seem to be trying to cache values in a map to speed up retrieval for a dynamic programming approach’.

In other words, it seems you’re already wrong. “Intent” is something 3.5 seems to handle and ChatGPT 4 is much better and 5, expected in September and, partially trained by its siblings for an ever faster evolution, even much, much better.

Statements about the limits of the capabilities of the current AIs is hazardous because they haven’t been sufficiently explored or understood, predicting their limits even in the near future seems foolish.