News for hackers. Stuff that matters.
Re:OpenAI, Google and Anthropic Are Struggling to Build More... by Xenoamorphous on Thursday November 14, 2024 @02:46PM (#42140363)
> Here is an example of a task that I do not believe this generation of LLMs can ever do but that is possible for a human

That’s possible for a highly intelligent, extensively trained, very small subset of humans.

Re:OpenAI, Google and Anthropic Are Struggling to Build More... by hatefulmoron on Thursday November 14, 2024 @03:34PM (#42140903)
If you took the intersection of every human's abilities you'd be left with a very unimpressive set.

That also ignores the fact that the small set of humans capable of building programming languages and compilers is a consequence of specialization and lack of interest. There are plenty of humans that are capable of learning how to do it. LLMs, on the other hand, are both specialized for the task and aren't lazy or uninterested.

Re:OpenAI, Google and Anthropic Are Struggling to Build More... by luckydata on Thursday November 14, 2024 @03:52PM (#42141088)
does it mean people that can build languages and compilers are not humans? What is the point you're trying to make?
Re:OpenAI, Google and Anthropic Are Struggling to Build More... by fragmede on Thursday November 14, 2024 @04:02PM (#42141178)
It means that's a really high bar for intelligence, human or otherwise. If AGI is "as good as a human, and the test is a trick task that most humans would fail at (especially considering the weasel requirement that it additionally has to be faster), why is that considered a reasonable bar for human-grade intelligence.