3 ms·

We have watched many humans grow so we have a pretty good idea of the curve. A better analogy is an alien blob appeared one day and went from 3kg to 9kg in a ye

We have watched many humans grow so we have a pretty good idea of the curve. A better analogy is an alien blob appeared one day and went from 3kg to 9kg in a year. We have never seen one of these before, so we don't know what it's growth curve looks like. But it keeps eating food and keeps getting bigger.

mlsu 1y ago
Mine's different. She's cuter. On a more serious note. Have these AI doom guys ever dealt with one of these cutting edge models on out of distribution data? They suck so so bad. There's only so much data available, the models have basically slurped it all. Let alone like the basic thermodynamics of it. There's only so much entropy out there in cyberspace to harvest, at some point you run into a wall and then you have to build real robots to go collect more in the real world. And how's that going for them? Also I can't help remarking: the metaphor you chose is science fiction.
- ben_w 1y ago
  > On a more serious note. Have these AI doom guys ever dealt with one of these cutting edge models on out of distribution data? They suck so so bad. There's only so much data available, the models have basically slurped it all. Many. Some of the doomers[0] even work at the big AI labs. Example of why this isn't sufficient to make them safe: Ebola is really stupid, infecting humans is effectively "out of distribution" for its genome, it is nevertheless really bad precisely because being out of distribution is why it often kills hosts too fast to infect enough other people to become a pandemic or endemic. > Let alone like the basic thermodynamics of it. There's only so much entropy out there in cyberspace to harvest, at some point you run into a wall and then you have to build real robots to go collect more in the real world. And how's that going for them? Apart from the existing few million industrial robots[1], the fleets of cars (not just Tesla but also) and farming equipment, all the drones including but not limited to delivery robots[2], and the robots widely used in scientific research to automate chemical and biological testing[3]? (But apart from that, what have the Romans ever done for us?) Don't get me wrong, I am willing to say that AI as it currently exists is "stupid" because of how much more training data it needs than most animals, let alone humans. It may be that any plausible number of robots in the real world still doesn't learn fast enough. And it may be that we're not smart enough to figure out how to make them as smart as us in any plausible timeframe, even though evolution managed it eventually without planning for us. But "you need robots" is not a strong argument. [0] for the purposes of this comment, P(doom)>0.1, though I am also critical of even just phrasing it in such a simplified way: https://benwheatley.github.io/blog/2024/11/18-13.16.17.html https://benwheatley.github.io/blog/2024/11/18-13.16.17.html [1] https://ifr.org/ifr-press-releases/news/record-of-4-million-robots-working-in-factories-worldwide https://ifr.org/ifr-press-releases/news/record-of-4-million-... [2] I assume this is basically just a Roomba with a fridge on it, but the sign on the elevator door is amusing: https://en.wikipedia.org/wiki/File:Hospital_delivery_robot_having_priority_to_elevators.jpg https://en.wikipedia.org/wiki/File:Hospital_delivery_robot_h... [3] https://en.wikipedia.org/wiki/Laboratory_automation https://en.wikipedia.org/wiki/Laboratory_automation
HarHarVeryFunny 1y ago
Yeah, but we're not talking about alien blobs, we're talking about pre-trained transformers. I'm 100% certain that if you make them bigger and better then all you will have is a bigger better pre-trained transformer. Scale it up (and sprinkle some magic fairy dust on it?) and it'll become sentient seems to be the thought process. Didn't work for CYC, and not going to work here either. We need architecture, not scale or efficiency or bells and whistles. Get rid of pre-training and design an architecture and learning algorithm that will learn continuously and incrementally from it's own actions and mistakes (i.e. prediction failures) and we'll start to get somewhere.
_Algernon_ 1y ago
We have seen a lot of things grow before though, so the pragmatic choice is to build some diminishing returns into the model. Not doing so is alarmist.

mlsu 1y ago

Mine's different. She's cuter. On a more serious note. Have these AI doom guys ever dealt with one of these cutting edge models on out of distribution data? They suck so so bad. There's only so much data available, the models have basically slurped it all. Let alone like the basic thermodynamics of it. There's only so much entropy out there in cyberspace to harvest, at some point you run into a wall and then you have to build real robots to go collect more in the real world. And how's that going for them? Also I can't help remarking: the metaphor you chose is science fiction.

ben_w 1y ago
> On a more serious note. Have these AI doom guys ever dealt with one of these cutting edge models on out of distribution data? They suck so so bad. There's only so much data available, the models have basically slurped it all. Many. Some of the doomers[0] even work at the big AI labs. Example of why this isn't sufficient to make them safe: Ebola is really stupid, infecting humans is effectively "out of distribution" for its genome, it is nevertheless really bad precisely because being out of distribution is why it often kills hosts too fast to infect enough other people to become a pandemic or endemic. > Let alone like the basic thermodynamics of it. There's only so much entropy out there in cyberspace to harvest, at some point you run into a wall and then you have to build real robots to go collect more in the real world. And how's that going for them? Apart from the existing few million industrial robots[1], the fleets of cars (not just Tesla but also) and farming equipment, all the drones including but not limited to delivery robots[2], and the robots widely used in scientific research to automate chemical and biological testing[3]? (But apart from that, what have the Romans ever done for us?) Don't get me wrong, I am willing to say that AI as it currently exists is "stupid" because of how much more training data it needs than most animals, let alone humans. It may be that any plausible number of robots in the real world still doesn't learn fast enough. And it may be that we're not smart enough to figure out how to make them as smart as us in any plausible timeframe, even though evolution managed it eventually without planning for us. But "you need robots" is not a strong argument. [0] for the purposes of this comment, P(doom)>0.1, though I am also critical of even just phrasing it in such a simplified way: https://benwheatley.github.io/blog/2024/11/18-13.16.17.html https://benwheatley.github.io/blog/2024/11/18-13.16.17.html [1] https://ifr.org/ifr-press-releases/news/record-of-4-million-robots-working-in-factories-worldwide https://ifr.org/ifr-press-releases/news/record-of-4-million-... [2] I assume this is basically just a Roomba with a fridge on it, but the sign on the elevator door is amusing: https://en.wikipedia.org/wiki/File:Hospital_delivery_robot_having_priority_to_elevators.jpg https://en.wikipedia.org/wiki/File:Hospital_delivery_robot_h... [3] https://en.wikipedia.org/wiki/Laboratory_automation https://en.wikipedia.org/wiki/Laboratory_automation

HarHarVeryFunny 1y ago

Yeah, but we're not talking about alien blobs, we're talking about pre-trained transformers. I'm 100% certain that if you make them bigger and better then all you will have is a bigger better pre-trained transformer. Scale it up (and sprinkle some magic fairy dust on it?) and it'll become sentient seems to be the thought process. Didn't work for CYC, and not going to work here either. We need architecture, not scale or efficiency or bells and whistles. Get rid of pre-training and design an architecture and learning algorithm that will learn continuously and incrementally from it's own actions and mistakes (i.e. prediction failures) and we'll start to get somewhere.

_Algernon_ 1y ago

We have seen a lot of things grow before though, so the pragmatic choice is to build some diminishing returns into the model. Not doing so is alarmist.