21 comments

[ 6.6 ms ] story [ 50.4 ms ] thread
"The concept that these AIs will take over the world is, with that absolute lack of human-like understanding of how the world works, impossible." Noam Chomsky
Hmm. Thinking not like a human would help it avoid and exploit the mental traps humans have. We might not even know they've taken over.
You might like Metal Gear Solid if this idea appeals to you.
It will have different taps, biases, errors. That much is already apparent.
That's not what Chomsky said. It's the author interpretation of what he said. Chomsky said: "That day may come, but its dawn is not yet breaking."
And what does Noam think a human does from the moment his eyes open and the amniotic fluid is drained from his ears? Humans gorge on hundreds of terabytes of data until they reach around puberty. Oh. But it's squishy chemical signals, not gold scratchings on glass. Oh well that's different.
I would say humans are pretty good at pattern matching, and thinking of my brain as a "statistical engine" has a lot of intuitive appeal. We are always piecing together an understanding of a scene based on imperfect data.
Every parent knows that human children learn to understand language and speak well before they’ve had time to get exposed to “hundreds of terabytes” of meaningful data. This obvious fact informs Chomsky’s work on language acquisition, which led him to posit an innate language capability.
A human experiments, predicts, learns and understands. A baby shoves everything in its mouth. It discovers how its actions affect the world, extrapolates, and experiments. It learns metaphor, learns how metaphor is useful to predict similar systems, learns the limitations of metaphor. It isn't just gorging data, it is creating data. It fills in the blanks with guesses and superstition, and maybe even uses reasoning to confirm it's speculation. A baby doesn't just regurgitate, no matter what a sleep deprived parent might think.
Noam Chomsky is fascinating to read if you're a 14 year old anarcho socialist like I was. His self-contradictory and confused opinions about the future highlight the fact that he's basically an ancient malcontent who's eked out his living trying to form an unsuccessful apparatus to do exactly what he claims to despise, i.e. lay out thought patterns for a group of followers who in his case never seem to materialize any power or wealth. (The secret is, they all consider him irrelevant after they get out of high school).
What generation of ChatGPT will generate ad hominem responses like this?
It's not an ad hominem to attack a public figure / celebrity on the basis of what they're famous for.

But to answer your question, no version of ChatGPT will ever be allowed to criticize anyone with any degree of disdain. It's a future-proof skill, maybe the last one reserved for sentient humans.

Ad hominem means attacking the person’s character or motivation rather than debating their arguments. Appealing to emotions rather than reason. That describes the comment I replied to.
this sounds like the ego of a 13 years old, with no power nor wealth.
Well if we're going to appeal to authority on AI, you can choose Chomsky...

or Altman, Hinton, Bengio, and Demis Hassabis...

I'd even say Chomsky's authority worth has been going down the drain lately
Can you elaborate on what his worst takes of late have been?
This was discussed at the time:

The False Promise of ChatGPT - https://news.ycombinator.com/item?id=35067619 - March 2023 (370 comments)

(Btw I don't think Chomsky wrote that op-ed. I'm sure it reflects his views, but there were two coauthors.)