I sort of agree with you, but the appropriate thing to do is to change the law, not to violate it in the name of "doing what's right". It's true that this would mean amending the constitution, and that's difficult, but they have the legal right to choose what they allow.

OTOH, it would be quite reasonable to deny that they are common carriers if they use editorial judgement as to what posts to allow. That would be an easier approach, and in line with what's been done in the past. I just feel that it's blatantly unconstitutional. (I think the Supreme Court disagrees with me, but that was the Warren court, perhaps the current one would agree...but probably not. That would limit the executive power.)

Yes. So it should be legal for them to make that change. It should also inform everyone that YouTube is even less trustworthy than you previously thought. (Unfortunately, it probably won't.)

There are lots of domains were physical evidence is either missing or impossible, yet where many people feel the need to have certainty.

Actually, the space is even larger than that. Every area of expertise implies an area that is not being examined, since people have only finite intelligence and finite time to explore. So...I "believe" in the EWG multi-world interpretation of quantum physics (with a few modifications). This is a belief, because I'm nowhere near expert enough in the field to have detailed knowledge. I *do* acknowledge that there are other interpretations that fit the existing data equally well, but I find them...distasteful.

Also, I believe that my wife was a wonderful woman. This is not based on globally accessible knowledge, partially because "wonderful" is not well-defined.

Etc.

You are wrong. AI has done mathematical proofs that were new. It *can* only be original by combining existing information into new patterns, but if the "rules of inference" are good, this can allow it to create something new and good.

OTOH, you are partially correct, in that it can't derive anything that wasn't already implicitly implied by the existing knowledge.,,because it can't currently run its own experiments.

N.B.: This is a comment about "AI" not about pure "LLM"s. Pure LLMs are a lot less reliable, because they've been designed to never admit that they are uncertain. And because they've been trained on the Internet.

What are the analogous problems in China?

During the early years the laser was called "a solution in search of a problem". Don't try to estimate what current AIs can do by the applications that they are currently shoe-horned into.

OTOH, every speculation as to how AI will develop further is *speculative*. That explicitly includes the speculation that it will not get any better or more efficient. (And I'd call the speculation that "we've reached top AI" at least as silly as "AGI will show up tomorrow and solve all our problems".

You oversimplify. I despise Rust, but it does address real problems. (I'm not sure how well, because I won't use it.) I'm thinking of thinks like deadlock, livelock, etc. As someone above pointed out, there are lots of applications that don't need to deal with that, and subsets can work for them. (The above poster worked in a domain where all memory could be pre-allocated.)

Rust felt like programming with one hand tied behind my back. So I dropped it. Only one reference to a given item it just too restrictive. Perhaps it is really Turing complete, but so is a Turing machine. But multi-threaded programs really do need a better approach. (My real beef with C++ (and C) though is their handling of unicode. So I'm currently experimenting with D [ https://dlang.org/ ], which seems pretty good for the current application (though honestly since it's I/O bound Python would be quite acceptable). )

You're wrong about potatoes. They *are* high in starch, but they aren't "nearly all" starch, like corn or wheat.

Yes, it's got a long way to go. Unfortunately, at least SOME of the changes are (currently) on an exponential growth curve, and people have very poor ability to project those. (And also at some point "limiting factors" will manifest, which aren't significant during the early part of the rise.)

There are quite plausible scenarios where we are still in the early part of the exponential growth curve. Nobody can prove whether this speculation is true or false, but we should be prepared in case it is true.

That kind of thing is something that centrally controlled economies are prone to. It's the mirror image is the problems experienced during the "Great Leap Forward". Market driven economies have different problems (monopolies, concentration of power in the hands of the greedy, etc.) . I'm not really sure which is inherently more deleterious. Perhaps it depends on details of implementation.

It's almost certainly "too late", but I wouldn't say that it's "too little", as I don't think it's even the right move. The announced goal, however (move manufacturing back to the US) is correct. OTOH, sabotaging world trade is a really bad move.

Yes, but the projections I've seen give them several years before their chips catch up (to TSMC). So the question is "Is this a worthwhile move *this* year?". Clearly they should consider the US an unreliable supplier, however.

Unfortunately, the data isn't consistent. That's why they need to make corrections. The question is "Do the corrections make it more nearly accurate?", and that's really hard to demonstrate. When there's too much noise in the signal, it's really difficult to filter it out without losing the signal.

When you're talking petabytes and "reading DNA", I don't think 60 minutes is the right order of magnitude.

I'm not sure this is going to be public facing. Many of the objections seem to assume that it is. OTOH, AI is known for returning the answers you want it to return, regardless of the truth of those answers. So perhaps it's the perfect yes-man.