3 ms·

It’s nothing unless it turns out to be something. I’ll draw an analogy to the LLMs and diffusion models that have lately rocked the computer world. These thing

It’s nothing unless it turns out to be something. I’ll draw an analogy to the LLMs and diffusion models that have lately rocked the computer world. These things are straight up science fiction. I would have said computers would have never been capable of art and poetry. It was all fantasy until it was suddenly reality. We’ll probably know if this can be replicated relatively soon. And if it can it will kick off a whole new branch of materials science and begin a multi-year race to commercialization. We are facing gigantic challenges in energy and climate. We need the win, so my fingers are crossed.

chasd00 3y ago
Dude I’m right there with you, fingers crossed!
dinvlad 3y ago
LLMs were/are massively overhyped though. Like it’s not even close to the sort of transformation they promised. Neat tech yes, world-changing solution to all of our problems? Very hardly, unless a very significant effort is spent (and yet undiscovered tech invented) to eliminate all of its significant practical problems. Academics (and similarly venture startup founders) like to sort of jump to conclusions along the lines of “discovery was the hardest part - now this thing could enable ... in the next X years” at the end of a paper. And then people spend many decades to actually reduce this to practice. So 90% of work belongs to applications, not technology in on itself.
- anonylizard 3y ago
  Its already overturned the entire art world. 1.2 years ago if you claimed that an AI will just come out and make artists mass panic you would be laughed at. Now art is somehow just 'solved'. Progress is continuous and massive from Midjourney v1 -> v5
  - dinvlad 3y ago
    Well, that’s exactly what I’m talking about - new unproven tech that created more problems that it solved. It didn’t “solve” art since that is meaningless without a human artist. Instead, it stole popular concepts and is regurgitating it without any originality or thought put into it, and then used “instead” of real artistical works by those who don’t really understand art, but just need “content”. In other words, merely copying/reproducing something is emphatically not art. The same thing happened with traditional (painting) art and photography - the latter didn’t really replace it but just supplanted it in the areas where one wasn’t really looking for art but for mere reproduction of reality. And then over time, photography itself has evolved into art, since people have learned how to channel their inner self through it, and aimed at quality instead of “quantity” (and unlike the former, the latter by itself became trivial). I don’t see the same happening with generative models quite yet (heck, they didn’t really replace photography either!)
- ricardobeat 3y ago
  I don’t think you realize what’s happening. In it’s 1+ year existence, it has already seeped into media production, art, design, architecture, engineering, pharma, agriculture, education and basically every field dealing with digital information. We don’t hear it but there is massive change happening underneath our feet.

chasd00 3y ago

Dude I’m right there with you, fingers crossed!

dinvlad 3y ago

LLMs were/are massively overhyped though. Like it’s not even close to the sort of transformation they promised. Neat tech yes, world-changing solution to all of our problems? Very hardly, unless a very significant effort is spent (and yet undiscovered tech invented) to eliminate all of its significant practical problems. Academics (and similarly venture startup founders) like to sort of jump to conclusions along the lines of “discovery was the hardest part - now this thing could enable ... in the next X years” at the end of a paper. And then people spend many decades to actually reduce this to practice. So 90% of work belongs to applications, not technology in on itself.

anonylizard 3y ago
Its already overturned the entire art world. 1.2 years ago if you claimed that an AI will just come out and make artists mass panic you would be laughed at. Now art is somehow just 'solved'. Progress is continuous and massive from Midjourney v1 -> v5
- dinvlad 3y ago
  Well, that’s exactly what I’m talking about - new unproven tech that created more problems that it solved. It didn’t “solve” art since that is meaningless without a human artist. Instead, it stole popular concepts and is regurgitating it without any originality or thought put into it, and then used “instead” of real artistical works by those who don’t really understand art, but just need “content”. In other words, merely copying/reproducing something is emphatically not art. The same thing happened with traditional (painting) art and photography - the latter didn’t really replace it but just supplanted it in the areas where one wasn’t really looking for art but for mere reproduction of reality. And then over time, photography itself has evolved into art, since people have learned how to channel their inner self through it, and aimed at quality instead of “quantity” (and unlike the former, the latter by itself became trivial). I don’t see the same happening with generative models quite yet (heck, they didn’t really replace photography either!)
ricardobeat 3y ago
I don’t think you realize what’s happening. In it’s 1+ year existence, it has already seeped into media production, art, design, architecture, engineering, pharma, agriculture, education and basically every field dealing with digital information. We don’t hear it but there is massive change happening underneath our feet.