An AI chip uses fuzzy math for human-like intuition

7
An AI chip uses fuzzy math for human-like intuition

There is much future for machine learning. One uses a black marker to dot and diagram the nodes of the human brain: the parts that are “ruminative, that think deeply, that ponder.

A startup is trying to approximate these neurons and synapses in its next-generation computer processors, which the company is betting can “mechanize intelligence.”

Artificial intelligence is indeed often thought of as complex software that mines vast datasets, but Knowles and his co-founder, chief executive officer Nigel Toon, do argue that more important obstacles still do exist in the computers that run the software.

The problem, that has been observed is that chips — known, depending on their function, as CPUs (central processing units) or GPUs (graphics processing units) — were not actually designed to “ponder” in any recognizable human way.

Whereas of course, human brains make use of intuition to simplify problems such as identifying an approaching friend, a computer might indeed try to analyze every pixel of that person’s face, comparing it to a database of billions of images before attempting to say hello. That precision, which made indeed made a lot of sense when computers were primarily calculators, is massively inefficient for AI, burning huge quantities of energy to process all the relevant data.

An AI chip uses fuzzy math for human-like intuitionWhen Knowles and the no doubt more business-minded Toon founded Graphcore in 2016, they put “less precise” computing at the heart of their chips, which they call intelligence processing units or IPUs.

The concepts in one’s brain are quite vague. It is really the aggregation of very approximate data points that do cause one to have precise thoughts. There are various theories on why human intelligence forms this way, but for machine learning systems, which do need to process huge and amorphous information structures known as “graphs,” building a chip that specializes in connecting node-like data points may prove key in the evolution of AI. Put another way, Graphcore is developing a brain for computers that, if its co-founders are right, will one be able to process information more like a human instead of faking it through massive feats of number crunching.

Simon Knowles, chief technology officer of Graphcore Ltd uses a black marker to dot and diagram the nodes of the human brain: the parts that are “ruminative, that think deeply, that ponder.” His startup is indeed trying to approximate these neurons and synapses in its next-generation computer processors, which the company is betting can “mechanize intelligence.”

Artificial intelligence is often regarded as complex software that mines vast datasets, but Knowles and his co-founder, Chief Executive Officer Nigel Toon, argue that more important obstacles still do exist in the computers that tend to run the software. The problem, they feel depends on their function, as CPUs (central processing units) or GPUs (graphics processing units)—were not designed to “ponder” in any recognizably human way.

Whereas human brains make use of intuition to simplify problems such as identifying an approaching friend, a computer might try to analyze every pixel of that person’s face, comparing it to a database of billions of images before rather attempting to say hello.

LEAVE A REPLY

Please enter your comment!
Please enter your name here