The current tech/IT sector is heavily relying on and riding hype trains. It’s a bit like the fashion industry that way. But this AI hype so far has only been somewhat useful.

Current general LLMs are decent for prototyping or example output to jump-start you into the general direction of your destination, but their output always needs supervision and most often it needs fixing. If you apply unreliable and constantly changing AI to everything, and completely throw out humans, just because it’s cheaper, then you’ll get vastly inferior results. You probably get faster results, but the results will have tons of errors which introduces tons of extra problems you never had before. I can see AI fully replacing some jobs in some specific areas where errors don’t matter much. But that’s about it. For all other jobs or purposes, AI will be an extra tool, nothing more, nothing less.

AI has its uses within specific domains, when trained only on domain-specific and fact-based data. You know, things like AlphaZero or AlphaGo. Or AIs revealing new methods not known before to reach the same goal. But these general AIs like ChatGPT which are trained on basically the whole web with all the crap in it… it’s never going to be truly great. And it’s also becoming worse over time, i.e. not improving much at all, because the web will be even fuller with AI-generated crap in the future. So the AIs slurp up all that crap too. The training data gets muddier over time.