That opinions should be divergent is what should be expected. Current AIs have "jagged capability". If what you want fits where they're good, you can be overly impressed. If it fits where they're poor, you can be overly skeptical.

Actually, people have "jagged capability" also. Don't ask me to win at football. But we have reasonable models for people. We don't have reasonable models for AIs. Making the problem worse, different AIs have different capability profiles. I don't know how good Claude is at coding, but it's reportedly a lot better than Gemini. OTOH, there are other areas where Gemini is better. So you've got to pick the right tool for the right job.

This confusion is what should be expected. I remember the early days of computer use. FORTRAN and COBOL were used in *different* areas. Each one was supreme within it's own particular area. And assembler was used in yet different areas. You couldn't just pick a computer, you had to pick the appropriate computer, get it configured correctly, and you were still likely to have problems.

A second may be too short, but I think an hour would be unreasonably extensive. Instead divide each minute into those four parts. IIUC, most trades are done by algorithms, not by humans. Somebody says something like "if stock A drops below this amount while stock B stays above that amount, buy (or sell)", and the computer implements the rule.

That would work, but there can be reasons why it's not best. I'm in favor of a per-trade tax that is higher for items held a shorter period of time, reaching 100% for trades sold within 10 nanoseconds. (And paid by the seller.)

You are proposing a specific method of action for the drug. It can't be true without a reason for being true. Identifying that reason can be quite difficult. Actually even proving "understimulated receptors" is difficult, though that's feasible, but to make it a useful explanation you need to demonstrate why they are understimulated.

I think it's now every 5-10 years. OTOH, it *does* depend on the extent of retraining required.

If HR implements those decisions, then HR is a fair target for blame. Your argument only means that one shouldn't ONLY blame HR, and that's definitely correct. Management deserves the blame for every mistake that it doesn't rapidly correct.

It's a lot more complex than even that.

Small colleges can generally pay more attention to individual students, but can't afford really expensive tools. So many fields can't really be taught at a small college. (Computing used to be that way. Small schools would mail the programs the students had written in to a central site which would run them and mail the results back. Not a good learning environment.)

I'm not certain, but how would you create those studies? First you need a theory to test against. E.g., why were those GLP-1 receptors under-stimulated? That would be an "underlying reason".

Actually, this is the kind of thing that can never really reach the bottom until you reduce it to quantum physics, but you may be able to reduce it to a higher layer than that which is still "almost universally true among chordates". But people prefer to use terms like "lazy" which are not well-defined, only analogically defined. (And they *are* useful as a macro-level analysis...but not at a more finely examined level.)

Actually, it does change habits. It tends to remove obsessive thoughts about food. It tends to break the habits of what you eat and when.

But that's not sufficient. Those habits appeared because of some underlying reason, which hasn't been addressed. And we don't know what those underlying reasons are. Your assumption of what doesn't change is not found in the experimental studies or therelevant case histories I've read about. So something else is going on.

The article may show nothing new, but the study is important. The result of the study was not surprising, but it needed to be done.

IIUC, it was suggested in public by medical professionals over a year ago that GLP-1 antagonists were a temporary measure, and cessation would result in a rebound. But it needed to be proven. (It probably still needs more proof, but I haven't been following it, and I'm a programmer, not a biologist or a medic.)

Note that their study was based on those using their chatbot. They've got a highly biased data sample.

I don't think that's quite accurate...but it's sure in the right ballpark. IIUC Anthropic tries to be "the good guys" and tries to be honest. But they *do* have a highly biased viewpoint.

Say you don't get "true AGI" (which by my definition humans don't have), but only twice the efficiency and scope of current AIs. (I'm NOT limiting this to LLMs, which are a subset of AIs.)

Then they will probably be able to to 75% more of the work, so, after job restructuring, you'll need 75% fewer people. This *will* make you more efficient (see Jevon's paradox) so more jobs will become available, but I really doubt that 3 times as many jobs as currently exist will be created. (Yeah, that's a lousy way to state it, since it measures "jobs" by "whatever would currently be a job", and that's the wrong measure. But I can't figure out what would be the right measure that would also be understandable.)

So say half as many people will be employed...or people will be employed half as much of the time. 20 hours/week sounds like an ideal solution, but not one we're likely to get to without a lot of social unrest. And different jobs will be automated/restructured at different times, so a legislated work week isn't a plausible answer.

Nah. As long as at least one person understands it, it counts as "human knowledge".

Otherwise the proof of the "four color theorem" would be when computers pushed beyond human knowledge. (That one was so long that no one human understood all of the proof.)

Define "true intelligence". In the case the search domain wasn't "stuff the people have done", but rather "stuff that can be validly derived from stuff people have done via valid mathematical operations". It basically needed to generate the area it was searching in. I'm not sure how much of what people do can't be expressed that way, if you replace "validly derived" by "guess is most likely".