Chatting with Mark Proctor just now. He's about to get Drools running on Collection-Oriented, Parallelized Rete - or something like that. However, in the process of working on my forecasting, I'm thinking of the analysis board in "The Majority Report" or "NCIS LA" - but in 3D then in multiple dimensions. Design work by 2012 or 2015 (OK, maybe 2020) will be done by modeling in 3D or 4D using holographic (or some other technology) so that we can move objects around NOT with our hands but with our mind. Hand-movement is WAY too slow for what you can think. And the models will involve time, space and multi-dimensions and massively parallel inputs.
Think about it: Your body has probably a million or more inputs to your brain all running at one time. All in parallel. What I need is a decent tool for real parallel work, not a single processor or even 128 processors, but a few million processors like in the old days of DARPA. Each one doing just a single, simple task. But all in parallel and all controlled by a massive Parallel CPU that can processes a million inputs at once. Real intelligence, real active agents or real-world forecasting will require just such tools.
And it might not be computers as we know them today. I suspect that these will be chemical simulations of the brain or maybe even real brain matter that is trained (something like a neural net but not exactly) such that we can interact with them using our own central cortex in some simple way - maybe like the "Matrix" but that was way too weird to work for what I'm thinking about. Actually, I DO know that the military is working with real brain matter to train it as a "computer" of sorts. I've seen it myself so it isn't just rumour.
Well, the headaches are back. Off to bed. But keep thinking about how to REALLY advance our thought process to model what we're really thinking and doing it quickly and easily. DARPA, TI, HP, Apple, CYC - SOMEBODY has to come up with a solution right away that is not hardware-based junk that we are being forced to use today.
SDG
jco
jco
Posts
1 comment:
It would be far more productive to license parallel RETE from Dr. Forgy, since he has spent over a decade on the problem. The wealth of knowledge Dr. Forgy has on the problem is far above everyone else. To reach the same point that Dr. Forgy reached today, someone would have to spend an equivalent amount of time. My bias 2 cents.
Post a Comment