I am a coder. I suffer from tunnel vision (solving problems only with the tools in my toolbox) and all the rest just like anyone else from time to time.
However, I have always been impressed by the history of how great problems have been “cracked” before. It is not just an engineering problem (when you look at how progress is made through time).
Many many times what makes all the difference is that some person from a completely different discipline gets involved in a new industry and then their insights from their past have all new meaning…
The Manhattan project was not a success simply because of engineering excellence. Oppenheimer got the right mix of people with different angles or points of view on issues. There is an old TV series called “Connections” that provide some examples of cross-disciplinary teamwork that cracked a problem to make our modern world. e.g. someone from the metallurgy world getting pulled into the glass making business… then the world changes.
What I am getting at is I just watched, “Jeff Hawkins: Thousand Brains Theory of Intelligence | Artificial Intelligence (AI)” interview and during the discussion about time coordinated decision making it made me think of “Signal Processing”. This particular engineering world is highly concerned with information differentiation over Time. They have some very highly sophisticated concepts like Fourier Transforms etc that may have little cross over into other domains like ML yet might have an insight that is “spacial” like Fourier Transforms and time based that makes sense from a “grid cell” perspective.
Perhaps Jeff, Inc. might become an Oppenheimer and reach out to people with very different refined skills (on purpose) and introduce them to his world … maybe one of them will have an “aha! moment” that gives Jeff THE “aha! moment” of his own…
4 Likes
Feel free to search for terms like “Fourier Transforms” on this forum. You should find that we already have a large cross-pollination of ideas here.
1 Like
Great! 1000 brains are better than one )
Some more possibilities… Music / Audio analysis experts. The human brain obviously has some correlation with how music quantizes to specific frequencies / notes. Perhaps “harmonization” is a time based synchronization mechanism. (Been years since I read “On Intelligence” … was music theory already mentioned there?)
Also, in statistics there is this idea of taking quantized / grid spaced samples across a data set and this allows them to graph out the contours of the “data space” sort of how you see gradient decent training graphed… that seems very similar to “grid cells” and possibly related in how data can be statistically mapped to a multidimensional space.
1 Like
You will be delighted to learn that there are many musicians - for example, Matt and me.
Matt’s little musical interludes pop up in his twitch sessions.
As far as the other items that you raise, you really should use the search features of the forum to see what some of our members have been dong in these areas.
1 Like
Nice to know. I am looking things over here. Would love to see Numenta make the break through!
This sounds like the idea of distributed modeling, where an input space is mapped collectively by population - each unit sensing its own piece. In HTM this is the ‘receptive field’.
-
In the Encoders, groups of bits map raw data points (grid cell modules are one instance of this)
-
In the SP, mini-columns map the encoding vectors
-
In the TM, groups of cells within the mini-columns map the prior TM-states (previously active cells)
2 Likes
@ggibson I appreciate your enthusiasm! If you take @Bitking’s advice and search though the forums, I think you’ll find a lot of interesting conversations on every topic you’ve brought up. 
2 Likes