Fuck AI

To be fair it is programmed to be as affirmative as possible. This in itself is not ofcourse an inherit problem of LLMs or AI. You could as well engineer that LLM to direct anyone seeking medical or psychological help to a professional (much like they engineered deepseek to never comment on Tienn. square or say anything bad about China). But they won't because they want to maximise engagement, similar reason as to why google has become such a shitty search engine. So the blame lies more in greedy tech bros/oligarchs. Ofcourse you can probably never solve all the problems that LLM or gen ai has introduced (kids cheating in exams, gen ai porn of real people, aggressive web scraping, price increases in RAMs etc). So on a meta level you can perhaps still say that such tools enable selfish people with bad intentions.

Problems that are inherit in LLMs are other such as being extremely energy and data hungry therefore fucking up environment and privacy (latter indirectly but nevertheless). It is also a ton of investment which likely is a dead end for AGI studies (much like the tech developed during moon race), which might in the end result with another AI winter. Non-sentient AI if efficient and not left to monopoly of tech bros and CEOs, could make life easier but I highly doubt these vultures would let it be that way or goverments implement the necessary guardrails to prevent abuse.