Hardware Thoughts?

General
1

My career began over 40 years ago as an embedded hardware designer at GM. Gradually, I moved more and more into the software realm, especially AI. Today, I’m feeling the pull of hardware again. There just seem to be some problems that are only efficiently solved at the transistor level. A CPU’s instruction set is far more determinative of success than is the O/S or high-level languages. I’m wondering what the future looks like through a Thousand Brains lens - is hardware a major issue, or is it just me?

The announcement today of Bezos’ massive investment in Tenstorrent is on my mind. (it’s a Toronto thing)

3 Likes
2

In short, I think hardware is a “when” question, and not an “if” question. For example, @vclay mentioned some hardware approaches in the David Eagleman Podcast with Jeff Hawkins - #8 by vclay comment.

2 Likes
3

The PDA race in the 1990s is informative. Creating something in software (function) with the intention of implementing the hardware to fit gave us the Apple Newton. Creating something entirely new (form) gave us the Palm Pilot.

1 Like
4

Yes, whilst I agree with the principles of the TBP (I came to similar conclusions regarding the need to learn from sensory input and interactions) I am minded to start from a different place. Object recognition is not where intelligence begins. For mammals, and other animals, it begins with building a mental model of your body from nerve feedback as you wiggle your limbs coupled with tactile, audio and visual maps of your environment. To this end I am trying to build a low cost robotic platform generating just such sensory data. The data can then be fed into something like the TBP structure, although I am more inclined towards a hierarchical neural network architecture.

1 Like
5

Each of us brings our own history and biases to the table. For some, it’s robotics. For some, it’s handheld and small computers (aka ‘intelligence amplifiers’). For some, it’s distributed and concurrent systems like Erlang. For some, it’s the nature* of computation itself. And for some, it’s human physiology and neurology.

The openness to all forms of thinking (intelligence) of the TBP is quite refreshing and wonderful. It’s evolutionary: try everything, keep what works.

  • Nature had computation long before we ever climbed down from the trees. Consider the ribosome.
2 Likes