I find any claim that superintelligence helps with physics to be a hoot.
Dark matter is the most notable contradiction in physics today, where there is a complete mismatch between the physics we see in the lab, in the solar system, and globular clusters and the physics we see at the galactic scale. Contrast that to Newton's unified treatment of gravity on Earth and the Solar System.
There is no lack of darkon candidates or MOND ideas [1] what is lacking is an experiment or observation that can confirm one or the other. Similarly, a 1000x bigger TeraKamiokande or GigaKATRIN could constrain proton decay or put some precision on the neutrino mass but both of these are basically blue-collar problems.
[1] I used to like MOND but the more I've looked at it the more I've adopted the mainstream view of "this dark matter has a galaxy in it" as opposed to "this galaxy has dark matter in it". MOND fever is driven by a revisionist history where dark matter was discovered by Vera Rubin, not Zwicky [2] and that privileges galactic rotation curves (which MOND does great at) over many other kinds of evidence for DM.
[2] ... which I'd love to believe since Rubin did her work at my Uni!
So the author can't conceive of a situation where a scientist can use AI to reduce his busy work by 4 hours a day? He's the one that sounds stupid here.
I think the author leaves out one important point which is that most people sound like idiots when put on the spot and asked to talk about things outside their core competency, and for these men that core competency is business. It's entirely possible they're bad at that as well but a priori you would probably expect them to do a lot better.
It's the human person Gell-Mann effect, we listen to CEOs talk about science, tech, and engineering and they sound like morons because we know these fields. But their audience is specifically people who don't-- and to them, thanks to the effect, they sound like they know what they're talking about.
I read the first few paragraphs but when I saw the size of my scrollbar, I decided the author is putting way too much thought and effort into owning a VC-funded tech CEO playing the game you play when you're running a a VC-funded tech company.
"Our product is the second coming of Christ and if you give me money now you'll 100000x your investment!" is the correct answer to all questions when you're in that position. I'm not saying it's admirable, but it's what you do to keep money coming in for the time being. It's not that deep.
Author makes a number of valid and valuable points, but desperately needed to edit this down for poignancy out of respect for everyone's time (including their own), taking their own advice ("clearly articulate what you're saying"). Don't make using an LLM to extract your point seem like such a good idea, eh?
I'm in the camp now that asks the question, if AI is so good, why are we still tethered to big tech? why hasn't an untrained human prompted a product out that is 10 times better than anything big tech has to offer. After all intelligence is free :-)
It's not even that. Why are trained humans who are now 10 times more productive haven't created new amazing operating systems, programming languages, game engines, browsers etc. We should have had a lot of outstanding products since AI hype started, but instead every company except the few doing foundational models seems to be just stuck and confused.
Ed Zitron's writing is inflamatory, but his points are extremely easy to grasp. I always find those reads very therapeutic, as looking at too much genAI discussion can make it seem like I'm the crazy one.
GenAI has all of the media attention in the world, all the capital in the world, and a huge amount of human resources put into it. I think (or at least hope) that this fact isn't controversial to anyone, be they for or against it. We can then ask ourselves if having models that can write an e-shop API really is an acceptable result, when looking at the near-incomprehensible amounts spent into it.
One could say "It has also led to advances in [other field of expertise]", but couldn't a fraction of that money have achieved greater results if invested directly in that field? To build actual specialized tools and structures? That's an unfalsifiable hypothetical, but reading Sam Altman's "Gentle Singularity[1]" blogpost, it seems like wild guesses are a perfectly fair arguing ground.
On a small tangent about "Gentle Singularity", I think it's not fair to scoff at Ed's delivery when Sam Altman also pulls a lot of sneaky tricks when addressing the public:
> being amazed that it can make live-saving medical diagnoses
The classification model that spots tumor has nothing to do with his product category, I find it very dishonest to sandwich this one example between to examples of generative AI.
> A lot more people will be able to create software, and art. But the world wants a lot more of both, and experts will probably still be much better than novices, as long as they embrace the new tools.
"the world wants a lot more of both" doesn't quite justify the flood of slop on every single art-sharing platform. That's like saying the world wants a lot more of communication to hand wave the 10s of spam calls you get each day. "As long as they embrace the new tools" is just parroting the "adapt or die Luddite!" argument. As the CEO of the world's foremost AI company I expect more than the average rage bait comment you'd see on a forum, the fact that it's somehow an "improvement" is taken for granted, even though Sam is talking about fields he's never even dabbled in.
The statement probably doesn't weigh much since my biases are transparent, but I believe there's just so much more intellectual honesty in Ed's arguments than in much of what Sam Altman says.
21 comments
[ 3.0 ms ] story [ 47.6 ms ] threadDark matter is the most notable contradiction in physics today, where there is a complete mismatch between the physics we see in the lab, in the solar system, and globular clusters and the physics we see at the galactic scale. Contrast that to Newton's unified treatment of gravity on Earth and the Solar System.
There is no lack of darkon candidates or MOND ideas [1] what is lacking is an experiment or observation that can confirm one or the other. Similarly, a 1000x bigger TeraKamiokande or GigaKATRIN could constrain proton decay or put some precision on the neutrino mass but both of these are basically blue-collar problems.
[1] I used to like MOND but the more I've looked at it the more I've adopted the mainstream view of "this dark matter has a galaxy in it" as opposed to "this galaxy has dark matter in it". MOND fever is driven by a revisionist history where dark matter was discovered by Vera Rubin, not Zwicky [2] and that privileges galactic rotation curves (which MOND does great at) over many other kinds of evidence for DM.
[2] ... which I'd love to believe since Rubin did her work at my Uni!
You give them too much street cred. I'm not convinced they're even good at that.
Teasing over the various Midwestern accents is sort of like dealing with boxing great Joe Louis: you can run, but you just can't hide.
It's the human person Gell-Mann effect, we listen to CEOs talk about science, tech, and engineering and they sound like morons because we know these fields. But their audience is specifically people who don't-- and to them, thanks to the effect, they sound like they know what they're talking about.
"Our product is the second coming of Christ and if you give me money now you'll 100000x your investment!" is the correct answer to all questions when you're in that position. I'm not saying it's admirable, but it's what you do to keep money coming in for the time being. It's not that deep.
GenAI has all of the media attention in the world, all the capital in the world, and a huge amount of human resources put into it. I think (or at least hope) that this fact isn't controversial to anyone, be they for or against it. We can then ask ourselves if having models that can write an e-shop API really is an acceptable result, when looking at the near-incomprehensible amounts spent into it.
One could say "It has also led to advances in [other field of expertise]", but couldn't a fraction of that money have achieved greater results if invested directly in that field? To build actual specialized tools and structures? That's an unfalsifiable hypothetical, but reading Sam Altman's "Gentle Singularity[1]" blogpost, it seems like wild guesses are a perfectly fair arguing ground.
On a small tangent about "Gentle Singularity", I think it's not fair to scoff at Ed's delivery when Sam Altman also pulls a lot of sneaky tricks when addressing the public:
> being amazed that it can make live-saving medical diagnoses
The classification model that spots tumor has nothing to do with his product category, I find it very dishonest to sandwich this one example between to examples of generative AI.
> A lot more people will be able to create software, and art. But the world wants a lot more of both, and experts will probably still be much better than novices, as long as they embrace the new tools.
"the world wants a lot more of both" doesn't quite justify the flood of slop on every single art-sharing platform. That's like saying the world wants a lot more of communication to hand wave the 10s of spam calls you get each day. "As long as they embrace the new tools" is just parroting the "adapt or die Luddite!" argument. As the CEO of the world's foremost AI company I expect more than the average rage bait comment you'd see on a forum, the fact that it's somehow an "improvement" is taken for granted, even though Sam is talking about fields he's never even dabbled in.
The statement probably doesn't weigh much since my biases are transparent, but I believe there's just so much more intellectual honesty in Ed's arguments than in much of what Sam Altman says.
[1] https://blog.samaltman.com/the-gentle-singularity