I agree even very well intentioned, honest people have just about everything telling them not to look to hard.

Consider you work for MSCRT and get the report from bug bounty. You confirm the issue, and you do the right thing and turn on the klaxons at MS.

After a little background check to confirm the reporter isnt likely a compromised person you look for 'obvious' signs this was exploited. Finding none, you report your initial results up the chain. Now your job is evaluated on closed incidents / reports at least in part. Your manager tells you wrap this one up close it out, because he knows everyone above all the way up to the C-suite, does not want this to be huge black eye.

Would you go on a phishing expedition in search of more tiny, easily disputed IOCs trying to sift back thru logs for a span of a year or more, knowing the really dangerous guys often have very long dwell times, or would you move on? If you found real proof of an issue you might be hero -or- motivated interests might try to discredit and vilify you, if you don't find anything you might be accused of violating instructions or even get into trouble for looking at logs and systems without an official cause..

there just isn't anyone even down to the front line engineers that really would *want* to find a problem if there was one. Just about everyone at levels at least in the near term has a better day if they 'see no evil'

No argument. These constraints are artificial, not something the system needs to work. Hence they can be circumvented. Makes the whole thing even worse though.

Yes. I do that for my CS students. I have never had to justify that even once, as they all understand what it is about.

That one is dangerous. You just nicely summed up what Adolf Hitler promised the Germans and what got him elected. And, to be fair, he delivered on those promises. For most. And with rather crass caveats. I am sure you did not intend that, but freedom always and critically includes freedom of mind and speech and access to knowledge and education.

Ah, yes. Animism. Not a sign of a strong mind. (That PhD is an indicator of, but not a sure sign of a smart person.)

As they are trained, general purpose LLMs are really just search engines with some aggregation and adaption capabilities. Obviously, no insight or understanding and obviously not a person in any way.

While I am not the one you directed your question at, I think people have always been bad at questioning authority. There is a number from sociology that says only about 10-15% of all people are independent thinkers and only about 20% (including the former) can be convinced by rational argument. The rest does not question authority, unless they are following one that questions another authority and that is something else. The problem we currently have (again) is that so many authorities are of really bad quality.

As to a dystopia, I would say "not yet". Information technology certainly gives the usual authoritarians tools like never before in human history. The reactions to that ranges from embrace (China, and lately the US), to real efforts to limit that (Europe, but with caveats). Depending on how that ends, we might get a dominance of surveillance states on this planet, and that is certainly dystopian. There are also strong and raising fascist tendencies (using the Wikipedia definition) and not a lot of awareness how bad that is.

Indeed. And if you do not at least try to do that, you are cheating them.

It is one of the reasons why monopolies are a massive problem and need to be prevented by regulation. But in the US, greed has long since taken over and the rest of the world is asleep at the wheel regarding this problem. Well, with the increasing unreliability if the US as a partner, maybe Europe will finally wake up now. Or not.

With regard to "model collapse" this is a very good point.

So new data will not be good at all. But old data will get worse as well. The aging is a massive problem. Put this two things together and what we currently see is a straw-fire in the process of burning itself out and no more straw is to be had.

My take is all that will survive is low-quality, outdated and "collapsed" LLMs for cheap or for free for the masses and small, really expensive, special purpose LLMs for some (not a lot) industrial and administrative purposes. Say, 5% of what the AI peddlers promised in actual impact. About the same as from the last AI hypes.

You are basically saying it is a lie with some cover story. That is also possible.

Give the extreme price and low production of Helium 3 on earth (it is basically only produced as Tritium decay product), "some other source" is just as fantastical as getting it from the moon.

Maybe a bit more expensive than you thought?

It is high time to regard events such as this as gross management negligence and to start holding the screw-ups personally responsible. Otherwise things will just continue to get worse.

It is even worse: The training data they got in their massive criminal piracy campaign _ages_ and becomes less and less valuable. Nobody is coming out of this as a winner, except for some of the scammers pushing the hype.

Nope. But they are making a determined effort to manipulate people into thinking that.

Good luck with that. DCs of this type are far too specialized for that to happen.

That one is not a surprise: The efficiency gains claimed are entirely fictional. What LLM-type AI is doing is to destroy value. And hence no boom.