On Wed, Jun 24, 2020 at 11:13 PM ducis <[email protected]> wrote: > Would you mind elaborating on how the respective communities are > >thinking about lossless compression as a solely or even primarily > automatic process >
The primary cognitive barrier I've encountered among AGI researchers to the use of lossless compression as THE model selection criterion among unified ecological models (human, such as sociology and natural such as climate change, etc.) is the confusion between the choice of model selection criterion and the choice of compression method. Understandably, AGI researchers focus on the latter (at least to the extent they accept Solomonoff Induction as the gold standard), so when they hear me appeal to lossless compression as the model selection criterion, they can't help but start thinking about what sort of algorithm would produce the executable archive without human-provided models. That's when they lose the forest for the trees, and go away thinking I don't know what I'm talking about. It's tragic. > and > >thinking about models of society and the environment as less than Turing > complete? > Statistics are based on Shannon Information. Solomnoff Induction is based on Algorithmic Information which, itself, is based on a Turing-complete range of algorithms. Statistics are inherently incapable of modeling a universe produced by a Turing-complete algorithm. Indeed, this is the whole point of Solomonoff Induction. While there is some controversy over the claim that the universe is produced by a Turing complete computation, that controversy ranges from Turing computation to super-Turing computation. To the best of my knowledge, no one is arguing that the universe's "programming language" is less than Turing complete (ie: less capable than a Chomsky Type-0 grammar <https://en.wikipedia.org/wiki/Chomsky_hierarchy#Type-0_grammars>). > > > > ------------------------------- > > > At 2020-06-25 02:29:11, "James Bowery" <[email protected]> wrote: > > My present motive for bringing up the "physics envy" trope, and Minsky's > redemption, is the chasm over which civilization is now passing the abyss > may be bridged if only people get over the idea that Algorithmic > Information Theory's model selection criterion is "just another information > criterion like BIC, AIC, etc.". You don't need artificial intelligence AT > ALL to bridge this chasm with AIT. All you need is for people to stop > yammering (and potentially shooting) at each other about their social > theories and accept that the smallest executable archive of the collected > data available to us about society and the environment, should be accepted > as embodying the best theory we currently have for guiding public policy. > > Both the AGI and the Statistics community can contribute to bridging this > gap by giving up on their respective unprincipled behaviors relating to > AIT: The AGI community has to STOP thinking about lossless compression as > a solely or even primarily automatic process and the statistics community > has to STOP thinking about models of society and the environment as less > than Turing complete. Only then can the two sides bridge the chasm so the > vast sums of money sitting in passive piles under the control of senile > institutions will pour into prize purses for incremental reductions in the > size of the executable archive of said comprehensive data collections. > > Failure to do this _will_ result in "rivers of blood". > > On Wed, Jun 24, 2020 at 9:53 AM James Bowery <[email protected]> wrote: > >> >> >> On Wed, Jun 24, 2020 at 6:03 AM John Rose <[email protected]> >> wrote: >> >>> There may have been physics envy then since the technological >>> convergence of AIT and QIT had yet to materialize. >>> >> >> The "physics envy" trope is an excuse for being unprincipled while >> occupying positions of trust, power, privilege, responsibility and/or >> authority. Minsky had a life-long habit of being unprincipled while >> occupying a position of trust, power, privilege, responsibility and/or >> authority. This goes back, at least, with the conclusion of the book >> "Perceptrons" which started the first connectionist winter. After my >> colleague, Charles Sinclare Smith, managed to end that first winter when he >> took control of the Systems Development Foundation's endowment and financed >> the early work of Hinton, Werbos, Rumelhart, MacLelland, etc. Minsky had >> the gaul to get up in front of the 2nd IJCNN -- a time when those of us >> struggling to rekindle neural network research had finally overcome >> Minsky's first connectionist winter -- and address us plenary regarding his >> unprincipled "society of mind". I was among those who walked out when he >> took the stage. Although AFAIK Minsky didn't have a hand in the second >> connectionist winter, his 2007 invocation of "physics envy" came at a time >> when it was ending based primarily on Hinton's exploitation of hardware >> advances. >> >> I see Minsky's 2014 admission as a deathbed confession. As far as I'm >> concerned, he redeemed himself at least to the extent that people he had >> misled throughout his life heard it. >> >> Epstein? Do we *really* need to go there? >> > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + delivery > options <https://agi.topicbox.com/groups/agi/subscription> Permalink > <https://agi.topicbox.com/groups/agi/T07361bd0216a4e97-M425f5f69cd405ae5f73a5005> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T07361bd0216a4e97-M25d5b7b89c88320bfe311167 Delivery options: https://agi.topicbox.com/groups/agi/subscription
