Big-in-Japan AI code 'Chainer' shows how Intel will gun for GPUs

Ever heard of “Chainer”, the open-source framework for creating neural networks? I hadn't either until yesterday Intel decided to give it a big hug, taking Chainer from being big in Japan, where its parent company Preferred Networks works with the likes of Toyota on secret projects, to rather greater prominence. Chainer can …

[John Smith 19]

At heart Intel makes hardware and will throw its superior linewidth at the problem. Until....

It runs out of linewidth.

Then it's a massive corporation whose used to a fat margin on high end chips sustained by it's ongoing bromance with MS, with the executive overhead to match.

You don't need neural networks to run Windows.

You don't need Windows to run core neural network applications.

And BTW has anyone repealed Amdah's law? I don't think so.

[shifty_powers]

Comparison crap

Given that by el reg''s own article, https://www.theregister.co.uk/2017/04/06/google_ai_chips_trounce_intel_nvidia/, TensorFlow is not designed for the training phase I am disappointed that they did not take that comparison to task as it is disingenuous to say the least.

[deive]

Hmmm, first I'd like to point out that machine learning is NOT AI.

Secondly, that may be why Intel bought Movidius and then seems to have killed their upcoming cheap USB Tensorflow accelerator :-( the bastards, I was looking forwards to hooking one up to a Pi.

[Mathman]

Neural Networks are a technique that allows software algorithms to appear to perform "intelligent" processing such as recognition tasks. This falls under the term "artificial intelligence".

Obviously there is more to AI than learning and NNs (or "deep learning").

Note that "artificial intelligence" never claimed to be something similar to human intelligence (or consciousness). Hence the term "artificial". It is just the study of systems that perform tasks that humans would regard as requiring some intelligence (even if none is needed. e.g. chatbots that blindly follow a simple set of rules or scrabble software that just brute force tries every possibility).

[John Smith 19]

A note about a lot of AI problems

They can be described as searching for a solution amongst a very large number of possible solutions.

So the challenge is to find a way to cut down a search space (which is usually exponentially growing) to something that can find a solution (ideally optimal but often just a solution) before the heat death of the Universe.