TPU - General Discussion

[jacob]

@Grumpy
A spinor is the simple answer. Sorry if I hadn't made the point before -- I thought I had...

Simple? Here's a citation from: http://www.sjsu.edu/faculty/watkins/spinor.htm

The concept of spinor is now important in theoretical physics but it is a difficult topic to gain acquaintance with. Spinors were defined by Elie Cartan, the French mathematician, in terms of three dimensional vectors whose components are complex.

You are right about the need for simple answers, but spinors are far from simple. Plus, they are out of context here because the inventor is not trying to confuse anyone with complex theory but rather to convey a picture of how the TPU operates. What we ought to be interested in is a theory that explains high power output in simple terms, that is easy to implement, and that is in agreement with TPU architecture and as much as possible with physics.

In the current context, spinors are however really useful to create a new spin and send everyone in the wrong direction.

[Grumpy]

yup...

[Grumpy]

yup...

[turbo]

how about,
clock-wise on the outside and counter-clock-wise on the inside?
well or the way around.
maybe i could better call it "horizontal" and "vertical" 
no no to me it still is counterclockwise and up.

M.

[Grumpy]

yup...