You missed the point. Small enough pieces are already understandable. But... Well... It's like the problem with global variables, a few are really useful, but when they multiply the code becomes unintelligible. There are too many interactions.

Whether we know how they work or not depends on how closely you look. If you look closely enough, every single decision is understandable. But it's so complex that nobody understands the ensemble. Even ordinary programs are sort of this way. The Obfuscated C Contest is just an extreme outlier, but any really large chunk of code written by somebody else (or by you over the last few years ... and not all at once) can be difficult to understand. As it gets bigger, and has more patches, the problem gets worse. These things are at an extreme edge of this problem. But every small enough piece of them can be understood...at least by somebody.

In this context inappropriate use of AI is a use that gets unchecked erroneous outputs that you care about or it more expensive than you want to pay.

OK, many people don't know how to use it yet.

AI is such a versatile tool that people keep trying to use it in inappropriate ways/places. Again, like computers in the 1960's & 1970's. I suspect that we're still at the point where most attempts will fail (some horribly), but some will be REALLY successful.

N.B.: Computers didn't become largely successful until structured programming and commodity software took over large parts of the market. That greatly reduced the errors. What's the equivalent for AI? We won't know until after robots become common. (LLMs are just touching the edge of what AIs can do.)

It is (or at least can be), among other things, a good search engine.

The problem is that people don't know how to use it yet. EXPECT most attempts to fail. Most early attempts to use computers failed.

Of course, with AI the problem's a bit worse because it keeps changing so rapidly. Currently the rule of thumb is "You can train an AI do do a good job if it's easy to verify that a proposed solution is correct.",

You already acknowledged the catch when you said "Or, at least, any new ones.". Small groups are really much more interested in the kind that were useful in WW I, because those are easier to make. Doing development for new weapons is dangerous to the developer.

If IQ were valid, it wouldn't be particularly low. Unfortunately, it's sensitive to culture and education, so you are probably right. But don't confuse it with potential intelligence. IQ is basically a measure of how well you could do in school with reasonable effort, so if you didn't pay attention last year, your IQ is lower.

That said, intelligence *is* real. It's just that IQ is a really bad measure of it.

Called standard by who? I guess we shouldn't call the Imperial units, though. But I've *never* heard them called "standard" before.

I really doubt that it's the largest single market. I suspect that's China. The US and India are probably up there in the running too.

I'm not sure. I can imagine software displaying ads that I would not consider malware. E,g, Firefox displays many ads on Slashdot, and only the ones that hide parts that I'm trying to read to I really consider malware.

Define "malware". This would seem to fit my definition. Yeah, it's an edge case, but it seems on the "malware" side of the line.

I think you're quite optimistic. I'd go with "Security can be vastly strengthened", but even BSD has had bugs.

Currently you are overrating AI. There are circumstances where it's the right choice, and circumstances where it isn't. But the point keeps changing.

Well, yes. Because it enables people to take precautions against it. But it sure doesn't prove it won't happen.

Note: It doesn't even have to be done with malicious intent. Lots of people just write really unexpected prompts. The more you have doing that, the more edge cases you have.