Brandon Fishback

@PDoomOrder1 I like how he makes up a metaphor for intelligence and simply asserts it as true without giving any reason to believe so.

This is just obviously not true. There may be some domains where greater intelligence yields diminishing returns on prediction, but when that happens it is usually because of the structure of the world, not because intelligence itself is close to some general upper bound. In things like weather forecasting, for example, the limits come from chaos, noise, incomplete information, and other physical constraints. A much smarter system might still only do somewhat better there, but that says something about the problem, not about cognition as such. As a claim about intelligence itself, though, this is deeply anthropocentric. Humans have tiny working memory, slow reasoning speed, low cognitive throughput, limited recall, and we constantly lose track of long chains of abstraction. We are also bottlenecked by the sheer time it takes to gather, read, and integrate information. A huge amount of human thought is not spent reasoning at all, but shuttling information into and out of a very small workspace. It is easy to imagine a system with enough working memory to hold something like the entire internet in active use at once and reason over it coherently. A system like that would not be just slightly smarter than a human. It would be so far beyond us that the comparison would barely make sense. That is especially clear in mathematics. Humans are nowhere near any plausible upper bound there. We spend absurd amounts of time on simple-to-state problems because our minds are narrow, slow, and easily overloaded. Mathematics is also a relatively recent and highly artificial human capability, so it should not be surprising that we are pretty bad at it all things considered. A mind that could hold vastly more structure in view at once, search much deeper, and form abstractions humans could never unwind over the course of a lifetime would be incomparably smarter than we are. And collective human effort does not really rescue the claim. A large group of humans is not one unified mind. It is a loose coalition of small minds with communication loss, coordination costs, duplicated effort, and severe bottlenecks in how quickly information can be transmitted and integrated. Honestly, if you picked ten million random American adults and had them race Terence Tao to prove the Riemann Hypothesis, and told me one side solved it first, I would lean toward Tao. Raw headcount is not the same thing as coherent intelligence.

@vmgdev1 @tobyordoxford @fchollet But you don’t know that AI being a hundred times smarter is impossible. You’re just asserting it. There is no known speed of light limit to intelligence. We’re not talking about anything close to omniscience.

""" One of the biggest misconceptions people have about intelligence is seeing it as some kind of unbounded scalar stat, like height. "Future AI will have 10,000 IQ", that sort of thing. Intelligence is a conversion ratio, with an optimality bound. Increasing intelligence is not so much like "making the tower taller", it's more like "making the ball rounder". """ Anyway, this is pretty clearly a pointless discussion.

One of the biggest misconceptions people have about intelligence is seeing it as some kind of unbounded scalar stat, like height. "Future AI will have 10,000 IQ", that sort of thing. Intelligence is a conversion ratio, with an optimality bound. Increasing intelligence is not so much like "making the tower taller", it's more like "making the ball rounder". At some point it's already pretty damn spherical and any improvement is marginal. Now of course smart humans aren't quite at the optimal bound yet on an individual level, and machines will have many advantages besides intelligence -- mostly the removal of biological bottlenecks: greater processing speed, unlimited working memory, unlimited memory with perfect recall... but these are mostly things humans can also access through externalized cognitive tools.

@T43736689 @phl43 @1730 I’m not arguing you should take LLMs at their word. But googling a question doesn’t always address your claims. It could give results for other claims. LLMs address your claim and then you can fact check it.

@Bfish94 @phl43 @1730 No it’s good at mimicking a sycophant but the actual quality of the data just isn’t there. I do not need a machine to talk to me.

The replies to this tweet are completely delusional. It's really weird to me the lengths at which people go to deny how useful AI can be and exaggerate its flaws.

@vmgdev1 @tobyordoxford @fchollet OP never made the claim you are arguing against so what’s the point?

I was extremely clear, my argument is even boring. *I* was not trying to make a point about the how close a group of humans are to "optimal" intelligence like Fchollet was. The point I was trying to make is that there is no possibility of magical unbounded omniscient intelligence.

@vmgdev1 @tobyordoxford @fchollet Maybe you’re just not being clear. Also, no one is making the argument that AI will inherently grow unbounded indefinitely. The argument is that there’s no reason to think intelligence will only be marginally more intelligent.

@Bfish94 @tobyordoxford @fchollet No, it does not. I replied to a specific claim about the ability for intelligence to grow unbounded. You read into my argument that I was claiming AI intelligence will not be much smarter than humans. It's interesting how many people are misreading it though.

@vmgdev1 @tobyordoxford @fchollet Your argument has the same structure.

@Bfish94 @tobyordoxford @fchollet Like the other response, you are arguing with a strawman.

@vmgdev1 @tobyordoxford @fchollet “If travel speed grew unbounded, it would be faster than the speed of light, which is impossible. Therefore future tech can’t go much faster than a rocket.”

@tobyordoxford @fchollet If intelligence could grow unbounded at a certain point it would be equivalent to omniscience which itself is impossible because it would violate causality.

@T43736689 @phl43 @1730 Because AI is very good at directly engaging with what you ask it.

@phl43 @1730 You ask the AI for fact checking with sources and then you check the sources to make sure it’s not lying to you. Why even bother with the AI step? Just use google, it won’t throw a bunch of made up bullshit at you

@realprsn4sure There should be more samurai/cowboy team ups in fiction.

The cowboy and the samurai are natural allies.

Originality isn’t everything but it counts for a lot!

Funny thing is there are plenty of frontiers on Earth but no one wants to live in them.

@lymanstoneky So what you’re saying is that God wants trillions of us.

Canonically, the demon population is fixed, and as human population rises, the demons/people ratio falls. Since demons are also canonically finite beings, this suggests higher population strains the tempting efforts of demons.

@sircalebhammer @darrag69420 @lporiginalg You got that jew hair.

@darrag69420 @lporiginalg Why does everyone think I’m Jewish lol

When did digital computers get better than humans at math calculations

@Great_Chumpion @constans We had a 100 years of evidence that the government is worse at making rational choices.

@constans This is 50% of the socialist critique of capitalism by the way. Economics is literally based on the assumption that people almost always make rational choices I don't think people, including business executives, almost always make rational choices actually

You’d think the age of streaming would have allowed more niche content that could rely on a few million watchers, but no— “popular” things with a solid audience still get canceled after 2 seasons.

@arvofart What’s the point of taking a picture? Just learn to paint.

The thing that confuses me about people who use AI in media is the question of why they’re making that thing in the first place. If you don’t want to write a book, why generate one? If you don’t want to paint, why are you advertising fake paintings?

I did not expect the best Chuck Norris story I’ve ever heard to involve Jessica Simpson, Denzel Washington, and Barney the dinosaur.

@iamjoshgreen @sugarbets_ @MrDanielBuck No, kids are drilled in how to do multiplication/division over and over again until they master it. Later they use simple calculators for other math.

@Bfish94 @sugarbets_ @MrDanielBuck We do let kids use calculators before they master math. That’s literally how it works. And they still learned math.

Any work sent home or completed on a computer will be done by AI and chatbots The only option is hand written, old-fashioned tests and essays now

@sugarbets_ @MrDanielBuck That’s like letting a kid use calculators before they can do basic math.

@MrDanielBuck It should shatter your worldview if your worldview included thinking you can detect AI tools yourself. Would we tell graphic design students that they can’t use Photoshop? We should be using AI in the classroom

@reset_by_peer That’s just dumb. I use it to get more done at work.

Almost everyone in the world views AI as a cheat-code that lets you get out of doing work. It’s a tool that lets you pull one over on your annoying boss or the mean professor. If you are seeing AI output, however, it means someone is trying to pull one over on you, which sucks.