Ask HN: Is there a mental task at which average humans can still beat computers? | Hacker News Reader

Ask HN: Is there a mental task at which average humans can still beat computers?

1 pointp-e-w10 comments

It's certainly not: Anything that can be solved easily by computation (obviously), information retrieval, memory, general knowledge, creative writing, poetry, drawing, composition, ...

What, if anything, is left?

Please keep in mind I'm asking about the average human, not the average HN user with a PhD in whatever.

View on HN

There's a philosopher name Hubert Dreyfus who has written a couple of books about the subject. In 1972 he published What Computers Can't Do: The Limits of Artificial Intelligence, a 3rd edition published in 1992 was titled What Computers Still Can't Do. He also co-wrote, with his brother Stuart, Mind Over Machine: The Power of Human Intuition and Expertise in the Era of the Computer

If you're not up for a whole book, there a 1965 paper titled "Alchemy and Artificial Intelligence" available at https://www.rand.org/content/dam/rand/pubs/papers/2006/P3244... that weighs in at only 98 pages.

I'm talking about today. I'm well aware that in 1965, in 1972, and even in 1992 there were plenty of tasks at which average humans were obliterating computers.

Allow me to gently suggest that you read Hubert's work and considered current LLMs in light of his exegesis.

My inquiry is about concrete, measurable tasks for which the question of "superiority" makes sense to ask. I'm not familiar with the works you are referencing, but your response gives the strong impression that I would find metaphysical arguments or unjustified assumptions about the human mind, which I've seen many times and am not interested in.

If possible, please name a single concrete task that, according to those articles, the average human today can demonstrably beat a computer at.

Dreaming.

Dreaming is not a "task", and certainly not one where performance comparisons make any sense.

I sense that discussion is headed towards "it doesn't count if a computer can't do it", in which case, the premises begs the question. Good day, sir.

Take two humans. What does it mean for one of them to be "better at dreaming" than the other?

> creative writing, poetry, drawing

Show me a concrete, quantitative measure to evaluate the quality of those activities or their outcomes. Use that to demonstrate that LLMs are better than the average human. But first, define "average" for a poet, writer, or artist.

Developing ai models

Reply on news.ycombinator.com