5 ms·

This is very refreshing to read - looking from the outside in, there has been a pretty clear problem with (at least the representation of) ML research along the

This is very refreshing to read - looking from the outside in, there has been a pretty clear problem with (at least the representation of) ML research along the lines highlighted here - ridiculous hype, papers again and again linked from the same researchers and institutions indicating potential cliqueyness, nepotism and celebrity worship. All very dispiriting, indicating real problems making progress in AGI beyond the clear lack of ideas. Of course, just because these toxicity claims are being made, it doesn't mean they are accurate, but they certainly ring true. If they are, it is good to know they are being talked about in some quarters.

foolmeonce 6y ago
> Of course, just because these toxicity claims are being made, it doesn't mean they are accurate, but they certainly ring true. In many respects I think it is more important that we have no idea what is true about any industry. When I dwell too much about my impression of any part of the tech field I usually end up demotivated to do anything that might connect me with the social group involved. I think there is a general shallowness and overcompensation in our industry that makes it impossible to figure out what real life is like with a group until you are way too invested and they have already hazed you.
zippy5 6y ago
I'm super curious why you were optimistic about AGI in the first place? it seem to me that a majority of the performance gains in ML are a result of using better hardware to run brute-force statistics with larger more complex models but the algorithms themselves have been improving at a nominal rate.
- mellosouls 6y ago
  I'm optimistic about AGI because I see no reason for it not to be implemented (though the time-frame is a different matter). Going by the hype articles (which may be unrepresentative), we just seem to be moving faster and faster on an impressively powerful, but AGI-irrelevant train along a machine "learning" railway track and although I suspect plenty of people on the train would like to get off, the drivers and momentum are making that very difficult, as indicated in the OP article. I'm completely optimistic about AGI, just think we are allowing the excitement of the advances in Artificial Unintelligence over the last few years erroneously dominate our thinking about it - at least in the sort of papers that turn up in tech-related feeds. Again, this may be unrepresentative of the top thinkers in computer science (machine-learning/whatever). My own (layman!) opinion is that the good ideas have and will continue to come from external (or intersecting) fields, philosophy, neuroscience, etc; not computer scientists raving about the power of DeepWhatever using cloud-enabled networks.
  - zippy5 6y ago
    Thanks for sharing! I totally agree with you that we seemed to focus a little too narrowly. If you haven't read it already, you might enjoy the book Range awesome look at the impact of interdisciplinary innovation
    - mellosouls 6y ago
      I assume you mean: Range: How Generalists Triumph in a Specialized World by David Epstein. I'll check it out - thank you.
baja_blast 6y ago
AGI is very far away from being a reality. I am skeptical that classical computers will ever achieve it. We will need to build radically different machines to accomplish AGI.

foolmeonce 6y ago

> Of course, just because these toxicity claims are being made, it doesn't mean they are accurate, but they certainly ring true. In many respects I think it is more important that we have no idea what is true about any industry. When I dwell too much about my impression of any part of the tech field I usually end up demotivated to do anything that might connect me with the social group involved. I think there is a general shallowness and overcompensation in our industry that makes it impossible to figure out what real life is like with a group until you are way too invested and they have already hazed you.

zippy5 6y ago

I'm super curious why you were optimistic about AGI in the first place? it seem to me that a majority of the performance gains in ML are a result of using better hardware to run brute-force statistics with larger more complex models but the algorithms themselves have been improving at a nominal rate.

mellosouls 6y ago
I'm optimistic about AGI because I see no reason for it not to be implemented (though the time-frame is a different matter). Going by the hype articles (which may be unrepresentative), we just seem to be moving faster and faster on an impressively powerful, but AGI-irrelevant train along a machine "learning" railway track and although I suspect plenty of people on the train would like to get off, the drivers and momentum are making that very difficult, as indicated in the OP article. I'm completely optimistic about AGI, just think we are allowing the excitement of the advances in Artificial Unintelligence over the last few years erroneously dominate our thinking about it - at least in the sort of papers that turn up in tech-related feeds. Again, this may be unrepresentative of the top thinkers in computer science (machine-learning/whatever). My own (layman!) opinion is that the good ideas have and will continue to come from external (or intersecting) fields, philosophy, neuroscience, etc; not computer scientists raving about the power of DeepWhatever using cloud-enabled networks.
- zippy5 6y ago
  Thanks for sharing! I totally agree with you that we seemed to focus a little too narrowly. If you haven't read it already, you might enjoy the book Range awesome look at the impact of interdisciplinary innovation
  - mellosouls 6y ago
    I assume you mean: Range: How Generalists Triumph in a Specialized World by David Epstein. I'll check it out - thank you.

baja_blast 6y ago

AGI is very far away from being a reality. I am skeptical that classical computers will ever achieve it. We will need to build radically different machines to accomplish AGI.