4 ms·

Every time I see someone make an illogical prediction about AI, it's because they don't understand the frame problem. Immense processing power is of no value,

Every time I see someone make an illogical prediction about AI, it's because they don't understand the frame problem. Immense processing power is of no value, if you have no capability of grounding and regrounding it adaptively in reality. That's the job our bodies and our societies provide, and AIs have no way to do it, except taking on a human body and joining human society, at which point an argument can be made they're not AIs any more, but human cyborgs.

yorwba 9y ago
While I agree that AIs need lots of experience with all kinds of situations to take all of them into account correctly, I don't think that a human body is necessary for that. It could be a network of surveillance cameras, or a fleet of sensor-stuffed cars or any other source of large numbers of observations. To predict the effect of an action, you don't need to be able to execute it yourself; or even if you can do it, there is no need to have done it; simply observing it is enough. There is no reason to suggest that AIs couldn't learn similarly through observation.
- erikpukinskis 9y ago
  I don't mean to be rude, but did you read the article? The reason AIs can't learn just from observation is the frame problem. You use your body to interpret what you see. Animals who are paralyzed from birth are blind, not because they have no eyes, but because they have no embodied experiences to allow them to interpret what they are seeing.
  - yorwba 9y ago
    In spite of these subtleties, a number of solutions to the technical frame problem now exist that are adequate for logic-based AI research. Although improvements and extensions continue to be found, it is fair to say that the dust has settled, and that the frame problem, in its technical guise, is more-or-less solved (Shanahan 1997; Lifschitz 2015). Yes, I read the article. > Animals who are paralyzed from birth are blind, not because they have no eyes, but because they have no embodied experiences to allow them to interpret what they are seeing. That's interesting. Where can I learn more about that?
    - erikpukinskis 9y ago
      https://io9.gizmodo.com/the-seriously-creepy-two-kitten-experiment-1442107174 https://io9.gizmodo.com/the-seriously-creepy-two-kitten-expe...

yorwba 9y ago

While I agree that AIs need lots of experience with all kinds of situations to take all of them into account correctly, I don't think that a human body is necessary for that. It could be a network of surveillance cameras, or a fleet of sensor-stuffed cars or any other source of large numbers of observations. To predict the effect of an action, you don't need to be able to execute it yourself; or even if you can do it, there is no need to have done it; simply observing it is enough. There is no reason to suggest that AIs couldn't learn similarly through observation.

erikpukinskis 9y ago
I don't mean to be rude, but did you read the article? The reason AIs can't learn just from observation is the frame problem. You use your body to interpret what you see. Animals who are paralyzed from birth are blind, not because they have no eyes, but because they have no embodied experiences to allow them to interpret what they are seeing.
- yorwba 9y ago
  In spite of these subtleties, a number of solutions to the technical frame problem now exist that are adequate for logic-based AI research. Although improvements and extensions continue to be found, it is fair to say that the dust has settled, and that the frame problem, in its technical guise, is more-or-less solved (Shanahan 1997; Lifschitz 2015). Yes, I read the article. > Animals who are paralyzed from birth are blind, not because they have no eyes, but because they have no embodied experiences to allow them to interpret what they are seeing. That's interesting. Where can I learn more about that?
  - erikpukinskis 9y ago
    https://io9.gizmodo.com/the-seriously-creepy-two-kitten-experiment-1442107174 https://io9.gizmodo.com/the-seriously-creepy-two-kitten-expe...