4 ms·

Fellow physicist-by-training. Got interested in "machine learning" and "artificial intelligence" a few years ago. As I delve into the roots of the field, I incr

Fellow physicist-by-training. Got interested in "machine learning" and "artificial intelligence" a few years ago. As I delve into the roots of the field, I increasingly feel that many interesting ideas seem to have been discarded without justification other than the field switching en masse to a different fad at some instant. Of late, I'm particularly fascinated by ideas in Cybernetics, but I don't see anyone else being interested in them right now. I think there might be interesting insights into both biological and artificial intelligence (with communication definitely being an important component; consciousness is a word I avoid till I can find a more concrete handle on it), but it's primarily based on interest/taste, rather than motivated goal chasing :-) (I'm more interested in concrete ideas from Wiener, Ashcroft, etc. rather than the fluff that came later as the word's usage got stretched beyond meaning, as is now happening with "AI")

dr_dshiv 7y ago
Yes to cybernetics. Its day will come again! If you go deep, it's a rabbit hole to discover why teleology in science is so controversial. Wiener said cybernetics started from the idea that Intelligence is all about decreasing local entropy.
_Microft 7y ago
Maybe Jürgen Schmidhuber's opinion on consciousness might be of interest? https://www.reddit.com/r/MachineLearning/comments/2xcyrl/i_am_j%C3%BCrgen_schmidhuber_ama/cp44iba/ https://www.reddit.com/r/MachineLearning/comments/2xcyrl/i_a...
carapace 7y ago
In re: cybernetics and psychology, Gregory Bateson said that the Neurolinguistic Programming folks achieved what he had been working towards, a cybernetic psychology. If you're interested, start with the very first books "Structure of Magic" vol I and II https://www.goodreads.com/book/show/964154.The_Structure_of_Magic_I https://www.goodreads.com/book/show/964154.The_Structure_of_... https://www.goodreads.com/book/show/821993.The_Structure_of_Magic_II https://www.goodreads.com/book/show/821993.The_Structure_of_... Essentially, by modelling therapists' verbal patterns using Transformational Grammar a formal theory of mind was evolved that has lead to sophisticated algorithms for various kinds of psychological change. Unfortunately, a lot of pseudo-science and "woo" has grown up around it, but please don't let that distract you, don't throw the baby out with the bathwater. - - - - In re: creating simple machines with intelligence, IMO cybernetics (i.e. what's in "Introduction to Cybernetics" by Ashby; I'm also not interested in fluff) is pretty obviously the way to go. Even the simple centrifugal governor "seems alive" in the words of early witnesses. Purposive behaviour is the domain of living systems, at least according to instinct and experience, eh? I think one could use cybernetics to inform the design of robotic systems using e.g. the kind of simple controllers David Wyland talks about at the HomeBrew Robotics Club in this video: https://archive.org/details/HBRobotics_Forth https://archive.org/details/HBRobotics_Forth He's not explicitly talking about cybernetics, of course, but I think he's got the right idea and cybernetics shows the way forward.

dr_dshiv 7y ago

Yes to cybernetics. Its day will come again! If you go deep, it's a rabbit hole to discover why teleology in science is so controversial. Wiener said cybernetics started from the idea that Intelligence is all about decreasing local entropy.

_Microft 7y ago

Maybe Jürgen Schmidhuber's opinion on consciousness might be of interest? https://www.reddit.com/r/MachineLearning/comments/2xcyrl/i_am_j%C3%BCrgen_schmidhuber_ama/cp44iba/ https://www.reddit.com/r/MachineLearning/comments/2xcyrl/i_a...

carapace 7y ago

In re: cybernetics and psychology, Gregory Bateson said that the Neurolinguistic Programming folks achieved what he had been working towards, a cybernetic psychology. If you're interested, start with the very first books "Structure of Magic" vol I and II https://www.goodreads.com/book/show/964154.The_Structure_of_Magic_I https://www.goodreads.com/book/show/964154.The_Structure_of_... https://www.goodreads.com/book/show/821993.The_Structure_of_Magic_II https://www.goodreads.com/book/show/821993.The_Structure_of_... Essentially, by modelling therapists' verbal patterns using Transformational Grammar a formal theory of mind was evolved that has lead to sophisticated algorithms for various kinds of psychological change. Unfortunately, a lot of pseudo-science and "woo" has grown up around it, but please don't let that distract you, don't throw the baby out with the bathwater. - - - - In re: creating simple machines with intelligence, IMO cybernetics (i.e. what's in "Introduction to Cybernetics" by Ashby; I'm also not interested in fluff) is pretty obviously the way to go. Even the simple centrifugal governor "seems alive" in the words of early witnesses. Purposive behaviour is the domain of living systems, at least according to instinct and experience, eh? I think one could use cybernetics to inform the design of robotic systems using e.g. the kind of simple controllers David Wyland talks about at the HomeBrew Robotics Club in this video: https://archive.org/details/HBRobotics_Forth https://archive.org/details/HBRobotics_Forth He's not explicitly talking about cybernetics, of course, but I think he's got the right idea and cybernetics shows the way forward.