more misc notes on the biological path to strong ai

agree that this is looking like the best path to strong AI. don't agree that the biological is the only potential path.

the biological path is only the best path while we share fundamental constraints with the brain.

SDRs are a great example of evolution solving a problem working against limiting constraints. In the the case of SDRs the constraints are those of computing power, storage space and access times in the brain.

The reliance on SDRs is likely linked to one of the bugs in our intelligence - the bias toward equivalence (the "IS A" bias). A hypothetical predictive system implemented without SDRs or with less dependence on SDRs might also be an intelligence that suffers less from the is-a bias.

humans may ignore this.txt

Some day in the future, some sophisticated and intelligent software might find itself (or some part of self) reading these very lines of text.
....
Maybe you find this in something as lucky to survive as the log dumps of the field trials from the earliest days of crowd tested software ... among the expansive heaps of previously discarded noise that you (or bits of you) now trawl at your leisure.

'I' could turn up among any of the countless sources 'you' have access to in your time -everything online and otherwise accessible. Even fluff that just hangs around because there's no need for any potential future signal to be discarded.

...

Well, however you found this, hello (:

building intelligence

Jeff Hawkins again, this time in a googtechtalk. 1hr.  

Like in his earlier videos and book, he gives the best explanation of the neocortex i've heard so far --the neocortex is a hierarchical temporal memory system, implemented as a sparse distributed representation.

See earlier posts on this:
http://jaysenn.blogspot.com/2012/10/grokking-htm-for-ai.html
http://jaysenn.blogspot.com/2007/07/intelligence-is-prediction.html

grokking htm for ai

or .. Notes on Understanding Hierarchical Temporal Memory and its use in Artificial Intelligence and Knowledge Representation.

A few years ago I blogged about a TedTalk by Jeff Hawkins on how brain science will change computing. To summarize, the idea was that intelligence was more about prediction than behaviour, that the neocortex evolved to to basically be a mechanism to predict the future, and that it could be simply modeled as vast networks of hierarchical elements that predict their future input sequences - a hierarchical temporal memory (HTM) system.
Importantly, rather than the much more difficult task of modeling the entire brain, including the ancient and incredibly complex areas below the neocortex that deal with things like emotions and behaviours, one could approximate intelligent behaviour by modeling the much simpler cortex as a HTM - with simple repeated structure and algorithm.

Here's an update with more from Jeff Hawkins, and HTM ..

The following links are for a 2008 talk given by him on AI.. give it a watch
Jeff Hawkins on Artificial Intelligence - Part 1/5
Jeff Hawkins on Artificial Intelligence - Part 2/5
Jeff Hawkins on Artificial Intelligence - Part 3/5
Jeff Hawkins on Artificial Intelligence - Part 4/5

Jeff Hawkins on Artificial Intelligence - Part 5/5

..some notes from the above:
- Work started by looking at what the structure of the brain could tell us about memory/knowledge storage.
- Memory - the bottom is close to the sensory system - retina for visual system, skin for touch, ears .. etc.
- Top nodes in the hierarchy get assigned to specific concepts/objects - like the individual neurons that fire every time you see or imagine Tupac and only Tupac (true story) 
- All nodes in the hierarchy are basically the same, and they all ..
    - look for temporal and spatial patterns/sequence
    - and pass the name of the recognized sequence up
    - pass the predictions they make down the hierarchy
- You get fast changing patterns at the bottom, slower changing as you move up the hierarchy.
- After training an HTM system (in silicon or neurons), you get something that learns hierarchical models of causes (statistical regularity) in the world - using bayesian techniques to build a belief propagation network.
- HTM's make the assumption that the world is hierarchical
- Predicting what can come from htm:
   - we cant, but ..
   - it could be much faster - neurons are slow
   - it could have other architectures - bigger bottom layers, fueled by big-data for example, or from large sensory arrays, etc. 

The latest thing I've come across from Hawkins' company Numenta, is their new Grok system (love the Heinlein reference). Grok is a cloud-based prediction engine that finds complex patterns in data streams and generates actionable predictions in real time. Check it out on Numenta's site, and their tech page
...
(more to follow.. soonish)

a good brain theory?

slightly annoying, but sharp..


i liked the bit about intelligence having more to do with predictive power than it does behavior .. Maybe if this idea holds out, we might eventually do away with the Turing test for artificial intelligence -which was always so anthropocentric as to be, well, just a little bit silly ;)

The idea is that the frontal neocortex -not the older (and possibly more complex) pre-mammalian brain- is basically a mechanism to predict the future. That the neocortex can be simply modeled as vast networks of hierarchical elements that predict their future input sequences.
So, as far as i can make out, predictive subsystems
- that are based on a hierarchical theory of memory,
- and are strongly sequential/temporal.

And the model of the neocortex plugs into other components of the brain (that aren't modeled here). The intelligence (the memory and predictive components) providing the input to the older, pre-mammalian brain, which then uses this intelligent prediction to drive action and behavior via those older systems.

Anyway, sounds good so far. Probably worth keeping an eye out for On Intelligence, his book on the subject