5 ms·

The processes seem pretty alike to me (as a neuroscientist and AI researcher). Things will only move on from here, the next generation of these tools won't use

The processes seem pretty alike to me (as a neuroscientist and AI researcher). Things will only move on from here, the next generation of these tools won't use a training set of 5B images and complicated month long training procedures, they will allow the "ingestion" of a style by you showing it a single instance once of a target image and it will immediately know the style (just like a human artist would). I'm not putting any weight here on what is good or bad for society, but relying on that humans somehow work in a completely different way from where AI is and is going is not going to help. I do think it will take longer for the AIs to know all about human contexts though, so the pairing of human AD + bulk-gen AI seems to me to be an obvious near-term tag team that's hard to beat.

cycomanic 4y ago
I find this take interesting. So would you also argue that saving an image into computer memory is the same as memorizing an image for a human? Those processes are viewed very different by the law, but if we anthropomorphize computers should we not view them the same? Also I wonder where you get the view that future ML systems will not require large amounts of learning? I don't see any development in current systems that would allow that, or do you mean you have a network trained on large amounts of data which can then adjust to a style from a single image? If that's the case we are still at the same question, how was the original model trained.
- sebzim4500 4y ago
  > Those processes are viewed very different by the law, but if we anthropomorphize computers should we not view them the same? Not only do I think the two processes are essentially the same, but I can't think of any laws in my jurisdiction (the UK) which actually distinguish between them. E.g. we are allowed to make copies of digital media for personal use.
  - pcthrowaway 4y ago
    I didn't realize it's OK to record a film in a theater in the UK
    - sebzim4500 4y ago
      I'm actually not sure about that. It's certainly against the terms of the theatre, but assuming you are only using the recording for personal use, I'm not sure if you would be breaking any laws.
- l33tman 4y ago
  For the second question, yeah exactly, as long as you've trained the rest of the system to a certain degree you can certainly do one-shot training on top of that now already for object recognition for example and you would be able to do it for style acquisition for diffusion models as well soon I think (you can already pretty quickly do overfit training on them, at home, in a couple of minutes with 10-20 images). Essentially this is what the brain does when you do oneshot learning of traffic signs or characters when learning a new alphabet etc. (yeah sometimes it's not that easy but still it's "theoretically" possible :). The rest of the recognition pipeline is so general that styles and objects etc are just a small icing on the cake to learn on top, you don't need to retrain all the areas of the brain when adding a roadsign to your driving skill set. But my point was that you could train the rest of the network on more general public data and not greg rutkowski. Hooray. Then someone shows it a single greg image and you're back to square 0.
haswell 4y ago
What I meant by the content/context of processes was that one is a biological process that includes all of the context and constraints of evolution, while the other is still ultimately a man-made machine, operating with an entirely different set of constraints, ultimately at the direction of other humans. If we could develop literal eyeballs that could look at these images and translate the information the way humans do, the resulting capability is still no more human-like (in the sense that it should be afforded some human-like status) than any other program IMO. If we achieved AGI tomorrow, we'd still need to have a conversation about what it is allowed to "see", because our current notions about humans seeing things are all based on the constraints of human capability. Most people understand that a surveillance camera seeing something and a human seeing something have very different implications. In the short term, it's a conflation that I'd argue makes us see less clearly about what these systems are/are not, and leads to some questionable conclusions. In the long term, it's a whole other ball of wax that will still require either new regulations or new ways of thinking.
- theRealMe 4y ago
  I’ll be honest, this sounds like you have made a decision on your stance and now you’re building false distinctions to reinforce your own bias. You said a lot of words, but I believe your argument comes down to “computers are super powered compared to humans doing the same thing”? Is that accurate? Because magnitude of ability, to me, makes no difference at all. It’s perfectly acceptable for a human to study the artwork of a specific person and then create their own works based on that style. Why wouldn’t it be the same for an automated process?
  - Maursault 4y ago
    I think you're both barking up the wrong tree. A person, and even an animal, possibly even a plant or members of other kingdoms and domains, sees. A computer does not see any more than a lens sees, or to the extreme, a computer can not see any more than an empty paper towel roll can see. The computer, lens and empty paper towel roll have no "I," no ego. In order to see, there must be something, or more accurately, someone, seeing. AI is just a complex program, which is ultimately an algorithm, and to be very simplistic, a recipe. A recipe can never be conscious, can never have a sense of self nor a sense of anything. Just because a photocopier can reproduce an image doesn't remotely mean that it or anything within it could ever see anything.

cycomanic 4y ago

I find this take interesting. So would you also argue that saving an image into computer memory is the same as memorizing an image for a human? Those processes are viewed very different by the law, but if we anthropomorphize computers should we not view them the same? Also I wonder where you get the view that future ML systems will not require large amounts of learning? I don't see any development in current systems that would allow that, or do you mean you have a network trained on large amounts of data which can then adjust to a style from a single image? If that's the case we are still at the same question, how was the original model trained.

sebzim4500 4y ago
> Those processes are viewed very different by the law, but if we anthropomorphize computers should we not view them the same? Not only do I think the two processes are essentially the same, but I can't think of any laws in my jurisdiction (the UK) which actually distinguish between them. E.g. we are allowed to make copies of digital media for personal use.
- pcthrowaway 4y ago
  I didn't realize it's OK to record a film in a theater in the UK
  - sebzim4500 4y ago
    I'm actually not sure about that. It's certainly against the terms of the theatre, but assuming you are only using the recording for personal use, I'm not sure if you would be breaking any laws.
l33tman 4y ago
For the second question, yeah exactly, as long as you've trained the rest of the system to a certain degree you can certainly do one-shot training on top of that now already for object recognition for example and you would be able to do it for style acquisition for diffusion models as well soon I think (you can already pretty quickly do overfit training on them, at home, in a couple of minutes with 10-20 images). Essentially this is what the brain does when you do oneshot learning of traffic signs or characters when learning a new alphabet etc. (yeah sometimes it's not that easy but still it's "theoretically" possible :). The rest of the recognition pipeline is so general that styles and objects etc are just a small icing on the cake to learn on top, you don't need to retrain all the areas of the brain when adding a roadsign to your driving skill set. But my point was that you could train the rest of the network on more general public data and not greg rutkowski. Hooray. Then someone shows it a single greg image and you're back to square 0.

haswell 4y ago

What I meant by the content/context of processes was that one is a biological process that includes all of the context and constraints of evolution, while the other is still ultimately a man-made machine, operating with an entirely different set of constraints, ultimately at the direction of other humans. If we could develop literal eyeballs that could look at these images and translate the information the way humans do, the resulting capability is still no more human-like (in the sense that it should be afforded some human-like status) than any other program IMO. If we achieved AGI tomorrow, we'd still need to have a conversation about what it is allowed to "see", because our current notions about humans seeing things are all based on the constraints of human capability. Most people understand that a surveillance camera seeing something and a human seeing something have very different implications. In the short term, it's a conflation that I'd argue makes us see less clearly about what these systems are/are not, and leads to some questionable conclusions. In the long term, it's a whole other ball of wax that will still require either new regulations or new ways of thinking.

theRealMe 4y ago
I’ll be honest, this sounds like you have made a decision on your stance and now you’re building false distinctions to reinforce your own bias. You said a lot of words, but I believe your argument comes down to “computers are super powered compared to humans doing the same thing”? Is that accurate? Because magnitude of ability, to me, makes no difference at all. It’s perfectly acceptable for a human to study the artwork of a specific person and then create their own works based on that style. Why wouldn’t it be the same for an automated process?
- Maursault 4y ago
  I think you're both barking up the wrong tree. A person, and even an animal, possibly even a plant or members of other kingdoms and domains, sees. A computer does not see any more than a lens sees, or to the extreme, a computer can not see any more than an empty paper towel roll can see. The computer, lens and empty paper towel roll have no "I," no ego. In order to see, there must be something, or more accurately, someone, seeing. AI is just a complex program, which is ultimately an algorithm, and to be very simplistic, a recipe. A recipe can never be conscious, can never have a sense of self nor a sense of anything. Just because a photocopier can reproduce an image doesn't remotely mean that it or anything within it could ever see anything.