5 ms·

MKBHD's review of the new Sora release: https://www.youtube.com/watch?v=OY2x0TyKzIQ https://www.youtube.com/watch?v=OY2x0TyKzIQ

laweijfmvo 2y ago
Love the callout of them definitely training on his own videos
- danpalmer 2y ago
  ...which they shouldn't have been able to get? I had thought that it was against the YouTube ToS? (my personal understanding, unrelated to my employer)
  - Havoc 2y ago
    AI companies don’t give a shit about ToS. Hell most of the big players actively ignored copyright entirely in bulk. See thousand upon thousands of pirated books in the pile dataset. And right after that news broke they “fixed” the problem by stopping to disclose training data sources. Thats why early models had papers eg Llama 1 listed this and now nobody does. It’s just an unspoken yet open secret now.
    - potamic 2y ago
      How did they get access to pirated books?
      - leobg 2y ago
        Anna’s archive has files specifically for training LLMs. But I’d guess the big players secured their share beforehand, by scraping those sites. I have zero proof, it’s just a guess.
  - paxys 2y ago
    If AI companies respected ToS there would be no AI
  - aprilthird2021 2y ago
    The companies are fairly brazen, at least internally, about just scraping whatever, wherever and not caring about ToS of any website. All they really care about is blocking "bad" data that might make the models racist or sexual, etc.
awongh 2y ago
Interesting to see how bad the physics/object permanence is. I wonder if combining this with a Genie 2 type model (Google's new "world model") would be the next step in refining it's capabilities.
- kranke155 2y ago
  Until these models can figure out physics, it seems to me they will be an interesting toy
  - andybak 2y ago
    They can figure out a fair bit of physics. It's not a "no physics" vs "physics" thing. Rather it's a "flawed and unreliable physics" thing. It's similar to the LLM hallucination problem. LLMs produce nonsense and untruths - but they are still useful in many domains.
    - Barrin92 2y ago
      It's a pretty binary thing in the sense that "bad physics" pretty quickly decoheres into no physics. I saw one of these models doing a Minecraft like simulation and it looked sort of okay but then water started to end up in impossible places and once it was there it kept spreading and you ended up in some lovecraftian horror dimension. Any useful physics simluation at least needs boundary conditions to hold and these models have no boundary conditions because they have no clear categories of anything.
  - kylehotchkiss 2y ago
    But they don't, they just understand pixel relationships (right?)
    - potatoman22 2y ago
      You can model a lot of basic physics through observing 1,000,000 videos
replwoacause 2y ago
I quit watching this guy after he filmed himself speeding a 100 mph through residential. Just another privileged YouTuber.

laweijfmvo 2y ago

Love the callout of them definitely training on his own videos

danpalmer 2y ago
...which they shouldn't have been able to get? I had thought that it was against the YouTube ToS? (my personal understanding, unrelated to my employer)
- Havoc 2y ago
  AI companies don’t give a shit about ToS. Hell most of the big players actively ignored copyright entirely in bulk. See thousand upon thousands of pirated books in the pile dataset. And right after that news broke they “fixed” the problem by stopping to disclose training data sources. Thats why early models had papers eg Llama 1 listed this and now nobody does. It’s just an unspoken yet open secret now.
  - potamic 2y ago
    How did they get access to pirated books?
    - leobg 2y ago
      Anna’s archive has files specifically for training LLMs. But I’d guess the big players secured their share beforehand, by scraping those sites. I have zero proof, it’s just a guess.
- paxys 2y ago
  If AI companies respected ToS there would be no AI
- aprilthird2021 2y ago
  The companies are fairly brazen, at least internally, about just scraping whatever, wherever and not caring about ToS of any website. All they really care about is blocking "bad" data that might make the models racist or sexual, etc.

awongh 2y ago

Interesting to see how bad the physics/object permanence is. I wonder if combining this with a Genie 2 type model (Google's new "world model") would be the next step in refining it's capabilities.

kranke155 2y ago
Until these models can figure out physics, it seems to me they will be an interesting toy
- andybak 2y ago
  They can figure out a fair bit of physics. It's not a "no physics" vs "physics" thing. Rather it's a "flawed and unreliable physics" thing. It's similar to the LLM hallucination problem. LLMs produce nonsense and untruths - but they are still useful in many domains.
  - Barrin92 2y ago
    It's a pretty binary thing in the sense that "bad physics" pretty quickly decoheres into no physics. I saw one of these models doing a Minecraft like simulation and it looked sort of okay but then water started to end up in impossible places and once it was there it kept spreading and you ended up in some lovecraftian horror dimension. Any useful physics simluation at least needs boundary conditions to hold and these models have no boundary conditions because they have no clear categories of anything.
- kylehotchkiss 2y ago
  But they don't, they just understand pixel relationships (right?)
  - potatoman22 2y ago
    You can model a lot of basic physics through observing 1,000,000 videos

replwoacause 2y ago

I quit watching this guy after he filmed himself speeding a 100 mph through residential. Just another privileged YouTuber.