AI and GPT-based co-pilots.

These things makes everything more difficult because one has to be a real expert to catch the bullshit inevitably introduced by the way (the algorithm) these services generate the sequences of whatever.

A model does not build up from the first principles and does not check or evaluate anything for correctness. It just generates based on some abstract distance or frequency-based heuristic with a bit of randomness, to impose as an expert.

Ordinary coder will not even notice a flaw or a problem because one has no idea that there could be a flaw.

It is just, literally, like copy-pasting code from public forums without understanding. The code that is not written by you bottom-up from the first principles is a random potential disaster.

And yes, a co-pilot may change a few symbols. It does not have any semantic (or of any kind) knowledge whatsoever. Everything is just some abstract distance heuristics, of not even vowels or morphemes.

The whole thing is literally an illusion of “intelligence” – an artifical imposer.

All this is not an emotional alarmism, triggered by something unfamiliar and disturbing. Unlike most of you I do understand the underlying algorithms and techniques. It generates very convincing bullshit.