On the diversity of (future) intelligent machines

sequence-memory
encoders

#1

Preamble:
My tooling is more sociology and philosophy than neuroscience or computer science, and my day to day pursuits currently more commercial than academic. That said, I like to think I “get” HTM (if not all the math). This is a first post from me, so apologies if I miss anything in the protocols about categorization, content or bad humour. @rhyolight please move this post wherever you see fit.

Starting Point: The Functional Requirements of all Intelligent Systems

The ‘Hawkins List’ gives this field a set of ‘required’ and ‘parameter’ components to play with. The Diversity of future intelligent (cortically constrained) machines would therefore likely result from the various combinations, dimensions and cocktails of these ‘required’ and ‘parameter’ components.

I should perhaps have titled my post: “In praise of the Interplanetary Construction Worker and the tireless Smart Mathematician” - referenced by @jhawkins in the presentation below.

Lots and lots of hard work to be done - specially on the encoders

A sense -> an encoder -> sends forth --> for what benefit
e.g. Light -> Retina -> SDRs … -> Vision (of things, movement, spaces)
e.g. Sound -> Cochlea -> SDRs … -> Hearing (of things, music, phonemes)
e.g. Math functions -> “XYZ” -> SDRs … -> Math Genius

Thought Experiment: I hope this isn’t copyright infringement or perceived as defamatory in any way, but why not approach someone like Stephen Wolfram to work on that ‘Math Function Encoder’ (“XYZ”) which will inform the “Mathematical Savant” (tenured from Day 1, working without breaks or administration duties, getting up to speed fast on the known achievements of math to date and then who knows …). HTM and Alpha might make good bed-fellows.

Why stop at the (biologically immediate inputs of) audible sound and visible light waves? Could we encode for the non-visible ranges of the electro-magnetic spectrum? For Infrared? For X-Ray? And could we perhaps encode for Gravity? (hang on … that’s maybe a bit too Interstellar)

Questions

  1. Are there any more published or unpublished resources available to read on this topic? In the Forum? Beyond this forum, which are informed by HTM theory and practises?
  2. Will @jhawkins or @subutai be elaborating further – and soon – on the prospects for “Interplanetary Contractor” or “Smart Mathematician”?
  3. Has anyone yet kicked off a Taxonomy of Fantastic Cortical Machines (and where to find them)?

References:
NTX: Jeff Hawkins on Machine Learning (Full Presentation

Post from @fergalbyrne: http://inbits.com/2014/08/implications-of-the-geospatial-encoder/

The Arena for accountable long bets: http://longbets.org/predictions/


#2

Thanks @Curious_Mike for the thoughtful post. I enjoy thinking through future pathways like this.

Not really, but I like to think about this a lot. I have been musing about writing a blog post about what HTM might evolve into over the next decade or so. Lots of ideas there.

I would not expect that. Too many research problems to focus on first before we put serious thought into it.


#3

thanks Matt for the answers. Please do go ahead and write that blog! Love to read it. Happy to bounce more ideas off the forum and yourself in terms of “possible future cortical machines”.