AI on a chip..

Seems to be getting some focus lately. Nvidia’s Volta, ARM’s Trillium, Intel with Nervana, Microsoft with Catapult and Google with TPU are the starting points for the silicon AI.

With comparisons to 80387, GPUs the AI in silicon seems to be suited for getting high speed analytics into edge devices if dedicated chips can help in reducing device footprint rather than using general purpose CPUs.

 

Advertisements

Deep learning and AI

In the keynote address at HIMSS17 last week, IBM CEO mentioned three goals that should drive efforts on Deep learning and AI solution development:

1 Augment people and not replace

2 How and who trained the data, what was Business value that drove that.

3 New Collar jobs, data driven workforce.

Sounds common for other areas too!