DARPA shells out $21m for IBM cat brain chip The US Defense Advanced Research Projects Agency is moving ahead with IBM in the third leg of its Synapse cat brain chip. That leaves one more leg, a tail, and nine lives to go. Because this is the military, the third leg of the Systems of Neuromorphic Adaptive Plastic Scalable Electronics (Synapse) project at DARPA is called …
Cat brain ...
A burning desire to:
Kill everything that moves.
Eat 10 times a day
Sleep 20 hours a day
Then again, if we had a Skynet of cat brains, there'd be plenty of time to plot a fight back, 20 hours a day in fact.
Either that or we'd all be dead within 1 nanosecond, so the cat can consider it a job well done and get some quality nap / cleaning time in.
Oh this crap argument again.
Yes, brains are really efficient at a small volume of tasks (pattern recognition mostly).
.. but there is one major downside of "grey fuzzy cells between your ears" - no one knows how to program them. You can build a cat brain, but can you teach it to do something useful in a useful amount of time? Even tiny artificial neural networks of only a few thousand neurons are impossible to design by hand, unless you use simple neural models which are effectively "AND" and "OR" logic functions, but that's not how a real brain works (it is unashamedly analog).
And given you are taking 150K cores to solve this "problem" you better hope you end up with a pretty bloody amazing solution. 150K cores to plot an optimal trajectory to that mouse seems like overkill to me ...
I agree. The fallacy that a computer is a thinking machine is ubiquitous. A computer that isn't deterministic is faulty or badly programmed. It might be possible to build electronics to simulate the human brain, but such a device wouldn't be a computer, any more than it would be a screwdriver or a steam engine.
......And given you are taking 150K cores to solve this "problem" you better hope you end up with a pretty bloody amazing solution. 150K cores to plot an optimal trajectory to that mouse seems like overkill to me .....
If you recall a simple calculator once filled a large laboratory.....
Typically you throw whatever you can at a theory to prove it/get it to work, then when you have proof/working model, it gets refined.
I mean look at the LHC, in 100 years perhaps you may have a higgs-boson powered application on the latest iphone :D
Researchers will come up with some toy concept, bat it around a bit, then go off to the next pretty stone on the beach.
What will end up in the robot is a Von Neuman machine... or really, a pile of them, but pretty conventionally connected. After all, researchers have no legs....
Now, how do I get a PAYING job simulating cat brains and not having to produce anything but a pdf of the results?? It is hard to make things that actually work, vs. writing cat brain simulations and papers that talk about talking about making things maybe work (if someone else does all the actual work to make them work).
..... that Unlock Doors in the Minds
Who doubts that surreal men are designed by smart pussy processing to be out of this world servers to their heavenly delights with immaculately selflessly shared and exquisitely enjoyed flights of mutually satisfying fancy. ......... to be constantly refined in the never ending search for love in perfection.
One of Venus's Fab Secrets.
...so the goal here is to create things that think. There isn't a "thinking" thing on the planet that doesn't get things wrong, throws a wobbly and doesn't get randy. I don't want my PC trying to hump my leg or bite it whilst at the same time letting loose the cats of war because it banged its chassis off off a table. Mind you I quite fancy the idea of rebooting the cat.
