conspiracy theories are much more fun than the extremely boring truth: LLMs are not anywhere close to AGI because doing computer is hard - ThreadSky

rincewind.run • 451 days ago

conspiracy theories are much more fun than the extremely boring truth: LLMs are not anywhere close to AGI because doing computer is hard

Reposted from Matt Novak

LLM tech is just a plagiarism machine that can't actually reason. But I think you'll start to see this rationalization more and more. They'll insist AI is only failing at basic tasks that require logic because it's actually being sabotaged by woke employees at Google or something.

Comments

coldbrewsoymilk.bsky.social•450 days ago

That is some high degree brain rot conspiracy theories.

ilisabat.bsky.social•451 days ago

It isn't that I think it is IMPOSSIBLE to discover AGI without understanding things like the hard problem of consciousness; it's just that I think it's really fucking unlikely.

wades.bsky.social•451 days ago

As for me, I do not think "understanding" is going to happen with respect to consciousness
There might someday be an AGI but it will not illuminate the mechanisms of consciousness

ilisabat.bsky.social•451 days ago

Also, I totally agree that understanding will not come from an AGI; I just think an AGI is way less likely to be feasible if we haven't figured consciousness out.

Or maybe I'm totally wrong and we can build brilliant robo p-zombies.

wades.bsky.social•450 days ago

No human programmer(s) will ever "write an AGI." If there ever is one it will the product of some kind of iterative evolutionary process that produces the same kind of incomprehensible rat's nest of intertwined ad-hoc-ery that natural selection did.

ilisabat.bsky.social•451 days ago

Big believer in the intractability of the hard problem? I can dig it.

I exist in a superposition¹ of standard Christian dualism and Buddhist anattā, but either way I think we're not solving the hard problem.

¹if you call it this you can't get cognitive dissonance; those are the rules

iamjake.bsky.social•451 days ago

"We believe any deceleration of AI will cost lives. Deaths that were preventable by the AI that was prevented from existing is a form of murder." — Marc Andreesen, https://a16z.com/the-techno-optimist-manifesto/

lyndztanica.bsky.social•450 days ago

my keyboarding teacher in 1997 told me that computers only do exactly what you tell them to and that woman's words almost 30 years later are the most relevant pieces of knowledge i took from high school

nora-hakase.bsky.social•451 days ago

It is so Goddamned hard to get this across to people. Either it's "but they call it AI, capitalists would never lie or mislead in order to sell things" or it's "I have learned just enough about computers to be completely wrong about this and not know I'm wrong".

Aggravating

bossett.social•451 days ago

I love that they've started with the presumption that "this AI is totally infallible, and therefore any failures are corruption". Would that I had that kind of faith in *anything* on a computer.

asheepie.bsky.social•451 days ago

But that's also not understanding how the current AI systems function at the very, very basic level. It's like watching a wheel rolling down one hill and then up another using its momentum, then putting a stationary wheel at the bottom of a hill and expecting it to roll up it.

rincewind.run•451 days ago

I don’t have that much fucking faith in the calculator app

nightpups.bsky.social•451 days ago

now I'm thinking about all the programming errors in calculator apps, like the one in pokemon games.

ogfattcatt.bsky.social•451 days ago

And they called me crazy for buying every abacus in existence… well who’s laughing now!

ravenonthill.bsky.social•451 days ago

Didn't I show you this?

http://shinycroak.blogspot.com/2024/02/the-general-intelligence-of-robots.html

nightpups.bsky.social•451 days ago

I clearly should eat and go to sleep, it took way to long to translate AGI from acronym. LLMs are closer to parrots than AGI

geekgrammy.com•450 days ago

Fancy autocomplete.

czedwards.bsky.social•451 days ago

And the billionaires deeply invested in them are the most motivated to spread lies and conspiracy theories about Evil Anti-Capitalists who don’t want to Accelerate straight into the brick wall right in front of us.

oregonrolleda20.bsky.social•451 days ago

If they burn down the Amazon rainforest it might create enough energy they need to power the processors for it to make a joke that sounds like it came Seinfeld

dbarawn.bsky.social•450 days ago

I don't actually think it's because doing computer is hard. I think it's because they don't actually want an AGI, so they're not actually trying. An actual AGI would have all the problems a human has anyway.

polyparadigm.bsky.social•450 days ago

Also because intelligence isn’t general.

Comments

Posting Rules

Reply