Went back and read the announcement relating to milestone1 more carefully. Seems as if milestone2 is a hardware emulation. No little robot car being controlled here my friends. It's just the same as milestone1 and uses the same API except it uses a FPGA board instead of a computer spiking cortical neuron model.
"The BrainChip team have advised that having achieved this major milestone, they are now moving toward a hardware emulation using SNAP technology. Under this scenario, they believe the learning process would take significantly less time and would further demonstrate the key attributes of the SNAP, being its speed of learning whilst consuming substantially less power.
Curiously the model applied by AZK and used to configure the network of seven sensory neurons and four motor neurons simulated on a CPU for milestone1 is ascribed to Eugene M. Izhikevichand and goes back to 2003.
The only video you would see would be the same as that for milestone1 but a lot faster. Quicker and far more boring visually. No IP to give away either. Just the same video as last time but not lasting as long. No wonder they don't show it.
They think they can sell licenses off this half software half hardware based platform. What ever happened to SnapSim, "a simulator for BrainChip neurons that potential clients can use to develop applications before accessing the BrainChip platform".
Have they ever mentioned SnapSim again?
No real little car, no real little track, just more of the same computer modelled inputs into their off the shelf FPGA board. When will we see the real Brainchip controlling a real device in a real world situation? How many millions of development capital before we see this is the real question and will anyone be willing to licence the technology to build a real world chip and how much would they be prepared to pay for this early stage at this early stage?
Eshmun
BRN Price at posting:
26.5¢ Sentiment: Sell Disclosure: Not Held
(20min delay)