6 ms·

When I first came across this project on HN (early last year), I was taken aback by how impossible the project looked and how smart were people working on this.

When I first came across this project on HN (early last year), I was taken aback by how impossible the project looked and how smart were people working on this. Despite seeing a few intelligent names behind the project, I subconsciously believed that this would at least take 5-10 years before a breakthrough. Today I sit with the same amazement, taken aback again, appreciating how ridiculously awesome this is. Congratulations to the winners and everyone involved!

qingcharles 3y ago
So many things that look insane are becoming a reality. You look at those scrolls burnt to a crisp and the idea of reading them is nonsense. The fact I have a computer writing flowery alt text descriptions of my photos with unnerving accuracy is something I would not have predicted for another 20 years. But, here we are...
- Keyframe 3y ago
  Right? Imagine trying to explain some of it to one of the ancients - so, you have this quartz sand, see?...
  - p-e-w 3y ago
    That would be like explaining a baking recipe by starting with protons and electrons. State-of-the-art machine learning architectures aren't actually that complex. Diffusion models and transformers can be explained to a bright high schooler. I'm sure Archimedes and Euclid would have no problem understanding them. What they might have a problem understanding (or even imagining) is the mind-boggling amount of computation required to make those systems do anything useful. Getting Llama to produce a single token of text takes more calculations than all of humanity did by hand during all of Classical Antiquity.
    - hnfong 3y ago
      I think the quartz sand metaphor was to illustrate how advanced our silicon-based technology has become, not just the ML parts. Imagine all the stuff... transistors, Turing/Von Neumann machines, lithography, theoretical computer science, OS and compilers, the Internet... and lastly there's modern day machine learning that builds on top of all the above. The base level stuff isn't exactly protons and electrons, but given the nanometer scale of our chips, it's not that far away from the truth, and we (humanity) has somehow built amazing stuff on top of that.
      - arlcode 3y ago
        The basic of concept how such a thing is powered (electricity) is so far removed from anything people did in ancient times that it would be hard to get them to understand that neither gods nor magic are invovled. Smart "intellectual" people would certainly be willing to challenge basically everything they assume about nature, but I don't think your run of the mill farmer would be able to do that.
        adastra22 3y ago
        Well we are wizards who speak in arcane languages to thinking rocks to convince them to do our bidding. We speak to golems.
        hattmall 3y ago
        Electricity isn't much different from water. The concept of a water mill has been around a long time. So understanding the functional ideas wouldn't be all that difficult. As far as Gods I think you will have a tough case minimizing their involvement even today.
        ahazred8ta 3y ago
        It's simple, really; we flatten the rock, write on it with fire and acid, put lightning inside it, teach it many things, and make it speak to us from afar. See? No magic at all.
    - robertlagrant 3y ago
      > That would be like explaining a baking recipe by starting with protons and electrons It would be like explaining a baking recipe just talking about wheat and flour and heat. The point is that from first principles, ML is a huge jump. From first principles, baking is not.
  - lynx23 3y ago
    Sufficiently advanced technology is indestinguishable from magic...
- lynx23 3y ago
  Kudos for caring about alt tags. Blind user here. While we are at it. I was also thinking how I could make good use of the new vision models. And after a while... https://github.com/mlang/tracktales https://github.com/mlang/tracktales The fact I have a computer generate spoken narration for my MPD playlist with descriptions of album art included just blows my mind. 2023 was indeed a fucking milestone.
  - qingcharles 3y ago
    That's awesome. GPT vision is unreal, to me. It's neat that you can finally get the computer to describe the album covers, something that no-one would have thought to bother with previously as it seems so trivial. A lot of album covers are surreal, so you're really going to put GPT to work there! My grandma was blind, and I just spent 6 months looking after a guy who was blind (but has now had surgery and beat me in the eye test at the doctor's!), so I think about blindness a lot when I design.
p-e-w 3y ago
It's all about incentives. $1 million is a lot of money. The vast majority of hard problems don't have much brainpower dedicated to them, because the bang/buck ratio doesn't work out. Machine learning, math, and adjacent fields already have many careers that pay very well, so getting top-notch experts to dedicate their attention to what might be a futile endeavor is difficult. And this isn't only about the monetary value itself, but also the fact that a large cash prize attached to a challenge boosts the prestige of finding a solution. Nobel Prizes come with about a million bucks on top of them, after all. I'm quite confident that if someone offered $100 million for deciphering the Voynich manuscript or Linear A, we'd have a solution within 3 years.
- CityOfThrowaway 3y ago
  I'm 90% sure the people that did this project did it because they got nerd sniped by it and got to hang out with nat while earning a reasonable salary
- BenFranklin100 3y ago
  Not to be argumentative, but $1M isn’t very much money, certainly not for a project of this scope. It’s a testament to the creativity, competence, and dedication of those involved they’ve gotten this far with such little funding. Hopefully their early success will attract more resources to this very worthy project.
  - fragmede 3y ago
    it's $700k divided three ways, too. $234k is well within FAANG compensation range, but you get to work on such an awesome project.
    - jobs_throwaway 3y ago
      FAANG compensation range but no benefits and 100x the risk Make it $8m or $12m and FAANG employees can actually start to justify working on it seriously from a money perspective

qingcharles 3y ago

So many things that look insane are becoming a reality. You look at those scrolls burnt to a crisp and the idea of reading them is nonsense. The fact I have a computer writing flowery alt text descriptions of my photos with unnerving accuracy is something I would not have predicted for another 20 years. But, here we are...

Keyframe 3y ago
Right? Imagine trying to explain some of it to one of the ancients - so, you have this quartz sand, see?...
- p-e-w 3y ago
  That would be like explaining a baking recipe by starting with protons and electrons. State-of-the-art machine learning architectures aren't actually that complex. Diffusion models and transformers can be explained to a bright high schooler. I'm sure Archimedes and Euclid would have no problem understanding them. What they might have a problem understanding (or even imagining) is the mind-boggling amount of computation required to make those systems do anything useful. Getting Llama to produce a single token of text takes more calculations than all of humanity did by hand during all of Classical Antiquity.
  - hnfong 3y ago
    I think the quartz sand metaphor was to illustrate how advanced our silicon-based technology has become, not just the ML parts. Imagine all the stuff... transistors, Turing/Von Neumann machines, lithography, theoretical computer science, OS and compilers, the Internet... and lastly there's modern day machine learning that builds on top of all the above. The base level stuff isn't exactly protons and electrons, but given the nanometer scale of our chips, it's not that far away from the truth, and we (humanity) has somehow built amazing stuff on top of that.
    - arlcode 3y ago
      The basic of concept how such a thing is powered (electricity) is so far removed from anything people did in ancient times that it would be hard to get them to understand that neither gods nor magic are invovled. Smart "intellectual" people would certainly be willing to challenge basically everything they assume about nature, but I don't think your run of the mill farmer would be able to do that.
      - adastra22 3y ago
        Well we are wizards who speak in arcane languages to thinking rocks to convince them to do our bidding. We speak to golems.
      - hattmall 3y ago
        Electricity isn't much different from water. The concept of a water mill has been around a long time. So understanding the functional ideas wouldn't be all that difficult. As far as Gods I think you will have a tough case minimizing their involvement even today.
      - ahazred8ta 3y ago
        It's simple, really; we flatten the rock, write on it with fire and acid, put lightning inside it, teach it many things, and make it speak to us from afar. See? No magic at all.
  - robertlagrant 3y ago
    > That would be like explaining a baking recipe by starting with protons and electrons It would be like explaining a baking recipe just talking about wheat and flour and heat. The point is that from first principles, ML is a huge jump. From first principles, baking is not.
- lynx23 3y ago
  Sufficiently advanced technology is indestinguishable from magic...
lynx23 3y ago
Kudos for caring about alt tags. Blind user here. While we are at it. I was also thinking how I could make good use of the new vision models. And after a while... https://github.com/mlang/tracktales https://github.com/mlang/tracktales The fact I have a computer generate spoken narration for my MPD playlist with descriptions of album art included just blows my mind. 2023 was indeed a fucking milestone.
- qingcharles 3y ago
  That's awesome. GPT vision is unreal, to me. It's neat that you can finally get the computer to describe the album covers, something that no-one would have thought to bother with previously as it seems so trivial. A lot of album covers are surreal, so you're really going to put GPT to work there! My grandma was blind, and I just spent 6 months looking after a guy who was blind (but has now had surgery and beat me in the eye test at the doctor's!), so I think about blindness a lot when I design.

p-e-w 3y ago

It's all about incentives. $1 million is a lot of money. The vast majority of hard problems don't have much brainpower dedicated to them, because the bang/buck ratio doesn't work out. Machine learning, math, and adjacent fields already have many careers that pay very well, so getting top-notch experts to dedicate their attention to what might be a futile endeavor is difficult. And this isn't only about the monetary value itself, but also the fact that a large cash prize attached to a challenge boosts the prestige of finding a solution. Nobel Prizes come with about a million bucks on top of them, after all. I'm quite confident that if someone offered $100 million for deciphering the Voynich manuscript or Linear A, we'd have a solution within 3 years.

CityOfThrowaway 3y ago
I'm 90% sure the people that did this project did it because they got nerd sniped by it and got to hang out with nat while earning a reasonable salary
BenFranklin100 3y ago
Not to be argumentative, but $1M isn’t very much money, certainly not for a project of this scope. It’s a testament to the creativity, competence, and dedication of those involved they’ve gotten this far with such little funding. Hopefully their early success will attract more resources to this very worthy project.
- fragmede 3y ago
  it's $700k divided three ways, too. $234k is well within FAANG compensation range, but you get to work on such an awesome project.
  - jobs_throwaway 3y ago
    FAANG compensation range but no benefits and 100x the risk Make it $8m or $12m and FAANG employees can actually start to justify working on it seriously from a money perspective