Artificial Intelligence in autonomous devices

Artificial Intelligence advances in all possible ways. Both on large and powerful computers capable of performing complex simulations, as well as on specific small processors that can operate autonomously, without connecting to any central computer.

The cake is huge and continues to grow, and there seems to be a piece for almost everyone.

The people of Google have seen this very clearly and, in addition to using large water-cooled TPU servers, they are also developing an environment called Coral that uses specific ASICs with the ability to perform many machine learning algorithms on their own, thereby you can achieve results without having to rely on large online servers.

Logically, its calculation capacity is not as high, but it allows AI to work locally: which is often more than necessary.

With Coral, the people of Google offer hardware and software means so that AI applications can be made on them and thus try to be used by almost any company.

Thus the possible practical applications can be almost infinite.

The idea seems very good to me, since AI concepts are being introduced in more fields, and the fact that someone (in this case Google) offers the physical means to do it is a way to help the development of this technology. And incidentally get a share of that market.

Being able to have the AI ​​computing power on a local device is important from the point of view of privacy and speed of response.

People’s privacy must be maintained by any AI system that helps perform a function, from the medical area, security, or other fields.

And as for speed, the most immediate example is that of an autonomous vehicle, since in that case the reaction times are very few milliseconds, and there is no risk that the response will be late to the vehicle. All the calculation must be done locally.

The more Coral develops in both hardware and software, the easier it will be for developers to use it in their new products, and consequently it could become a commonly used “tool”.

I suppose that is what Google intends, but some companies may be afraid of falling into the hands of a very big one like Google.