The person you answered to has nothing, especially not any understanding of what AI research took place in the last decades. Hence they just try to confuse the issue and introduce something completely irrelevant as the new main topic. Obvious and unsophisticated.

The reality is that the LLM fans are essentially cultists and that they have completely lost sight of all the limitations we find in LLMs. Sure, LLMs are not useless. They allow better search (with sometimes really bad hallucinations and misinformation), they can summarize things in a way that helps experts (non-experts may be grossly misled though). What they cannot do competently is synthesize anything. Text-slop, code-slop, code that on repeated changes re-introduces regressions (which competent coders will not do), and even art- and music-slop. So, to better deal with the mountain of data we have today, they are somewhat helpful to actual domain experts, but that is it.

But the LLM fans think that the synthesis aspect is pretty good as well. It is not, at least to anybody that is a domain expert. My guess would be these people cannot perform on professional level either and hence think that slop is just fine. For example, the only mostly adequate code that LLMs can write is attack code from patches or patch-notes, because that is use-once, throw-away code. For anything else, as long as it is not beginner's level, an actual expert has to request in detail what is needed and make sure it is all there. At realistic token prices, that seems to be a lot more expensive than to have an expert write the code and that does not include the cost of low quality aspects that are not immediately obvious, like LLM code being long, hard to read and hard to review.

There is no reason to expect drastic cost reductions either. The current LLM business models need growth-rates they cannot get in the real world. Most of the current providers will crash and vanish. The rest will drastically reduce what they offer or charge prices that make most uses for synthesis non-viable.

Add that we will very likely find other serious drawbacks of having LLMs write things. This tech is on "demo" level currently and the demo was scaled up beyond all reason. Maybe it will deliver something reasonable in the synthesis use-case in 20 years, but that is not assured at all and not only because the inevitable crash long before.

Indeed. The main theory I have is this is an experiment how many utterly dumb people and how many complete assholes you can mix in without the whole thing becoming dysfunctional. I am not sure whether this is a success or a failure so far. I am tending towards "failure".

... without any or any good justification, this could work. Or not, if Bilibili turns out to be worse. In any case, YT has overstayed its welcome a while ago.

Naa. People are generally pretty dumb without social media. All TikTok will do is stop them from more failure using the pathetic excise for rationality the own.

Indeed. But the morons praying to the tech-god are not rational. They do not see what the actual reality is. And the are aggressively attacking anybody that calls out their delusion. Nothing new. Still utterly pathetic.

You are an idiot. The Space Nazi is one too. Data centers in space will not work.

No. Seriously. Think again about what you wrote. Here is a hint: An OpEd piece is not learning material.

Yes. But these exceptionally obvious facts are apparently too complex for the average person to understand.

Yep. Looks like people that are missing the very basics get admitted to university. That will not go well for society. Or for them.

Indeed. But this extremely advanced reasoning technique is far outside of the capabilities of people writing articles as stupid as the one we have here.

This article basically insinuates that using an LLM in one domain taints the work of a person in some completely different domain. Or something. This makes absolutely no sense. I have no idea how people can be so utterly mentally dysfunctional.

First, writing an OpEd piece is not teaching or learning math. An LLM could have told you that. Second, this person already knows math and is good at it. Hence using an LLM does not impact their skills.

Indeed. But somebody here just told me "copyright law treats humans and machines in the same way". That is complete disconnect and deep delusions. The LLM fans are operating on the level of a particularly stupid and clueless cult.

The stupidity on display by some people here, like you, is staggering. And then you push extreme lies on top of that.

You think using the _jargon_ implies competence? That is pathetic on another level. Yes, they use the term "learning", but they do understand it to NOT mean human learning.

Funny and pathetic. I do know enough to be not impressed. A CS PhD does that. Clearly you are a clueless fanboi with no insight.