Indeed. Enshittification is a fact of life for enterprises. But on the level of a whole society it is not sustainable.

Does not work. Sometimes you can actually make somebody understand, but generally people just assume they are smart without actual evidence.

So, yes, what is currently done and planned to be done with the misnamed "AI" is stupid. Yes, it will do a lot of damage. No, you cannot stop it.

Ype, that _is_ the MS business model. Lock people in as much as they can, because they sure cannot compete on quality.

That works only for 3rd rated and maybe 2nd rated people. If you have any dependency on a higher level of skill, you are screwed with that approach. But yes, that is a valid argument as many "experts" are not actually that good.

Well, maybe. A CEO needs agency, even if a complete asshole and idiot. AI does not have that. Maye to emulate your typical CEO just throwing facts about the company in there in an automated way would be enough though.

Primarily, I am still an IT security expert. Look at what happened, say, to Microsoft in the last 18 months. Code quality will be a question of survival in the not too distant future.

Ah, you are one of _those_! Hahaha, no, there is no indication either that should AGI be possible and we find out how to do it, that the result would be smarter than an average human, i/e/ still pretty dumb. Also, we have some very smart people in the human race. Do you see them get put into power? NO, they are mainly seen as a problem and a nuisance. You think that would be different with AGI?

1) We do not need it
2) Nobody knows whether it is even possible to create AGI. Not in the next few decades certainly as we do not even have theory how to do it. Remember that what LLMs do is based on theory that is more than 50 years old.

As to coders: No. AI cannot replace coders. A studente in my coding class summed it up nicely to me just recently: For a coding problem AI gives him 10 answers, 9 of which are well-sounding crap and one which is somewhat reasonable. This is for _very_ _easy_ stuff, taught in the first year. Now, if I do a simple web-search for the same question, I get about the same, but with fewer bad results and the explanations of the original authors attached. I prefer that and I always learn some things on the side when I do that. Hence AI does not even make coding faster, really. If you are _smart_ and can filter those 9 BS answers out, it may reduce what you need to know to code competently a bit. But that is it.

Well, I fully agree that current "AI" does just barely meet the requirements for "automation" (it is not really reliable enough for it), but the problem is that a lot of the traditionally needed calibration and adjustment work for automation has gotten massively easier due to a much improved NLP interface. That means a lot of jobs are going away, probably more than society can survive.

True. My take would be that no-clue, no-insight AI with specific, restricted training may be able to replace up to 20% of all workers. Not like full replacements, but things like where you keep 2 people of a team of 10 for the things AI cannot handle and sack the other 8. 20% job-loss with no replacements in a short time is on the level that can definitely destroy a society. Personally, I have not even applied for a job in about 25 years, I always got asked so I have no fear for me. But I am in no way representative and the average joe and jane do deserve to live decently too and must usually be able to do that. Society critically depends on most of its members doing reasonably well. As soon as that is not the case anymore, society slowly disintegrates.

Haha, no. Pure pity on my part. You are trying sooo hard! Sadly, trying does not lead to good results. Consider this your participation trophy.

You should get that functional analphabetism fixed. As long as you cannot even read, how is anybody ever going to take you seriously?

Because AI cannot do anything that needs understanding and if you sack the people that are doing that part it will become a massive problem and may even kill your organization.

The models are getting dumber because they try to "fix" them. That is essentially impossible without doing real damage.