The Future is spelt TPU

We are dropping Nvidia from portfolios and replacing it with Alphabet.

The story of the moment is the dark world of trading in GPU.

Where are all the GPU? Where did they go? Who bought them? Who profited?

Sanctions will do this. The USA treats sanctions as law, when they have zero force in law outside of the United States. They can only be enforced by coercion.

This man clearly responds to carrots and not sticks.

It is called a “Spy Scandal”. I think it was about a sale that benefited both parties.

There is an attitude that only the USA can have Artificial General Intelligence (AGI). Good luck with that one, since I do not believe it is even close to possible.

The issue is what LLM do well versus what they do poorly.

This is what an LLM is:

Output = Function(Input)

This is what a human is:

Output = Function(Input|Conscious State)

The state of that “Conscious State” makes all the difference.

You can fit functions to slowly changing patterns in data using a universal function approximator. That is what stacked neural networks are, and provably so.

Whatever bits of that unknown state are sublime is best left to spiritual authorities.

Human creative response to change is no stable pattern in tumultuous times.

Silicon Valley, and technology, are about money and not our collective spiritual health.

For some time, it had been believed that there was more money to be made boosting the idea that “AGI is just around the corner” and we can all be replaced by robots.

Commonsense was replaced by the short video musings of grifters:

The thing I am waiting for is that moment when the singularity disgorges hoarded ram.

The 40% of global DRAM production hoarded by one firm will emerge from the singularity at the center of that soulless firm called OpenAI.

This is commerce.

Moral principles and a higher calling never visited that computing campus.

Only foolish investors believe this hype-torrent of life in the singular hereafter.

The correct response is to lower your sights on AGI, and execute a quick table turn.

Now that Anthropic has a spy, whom you pay, called Claude, embedded deep within the software bowels of most every major enterprise, the tables can be turned.

Claude is good at coding and works for a firm called Anthropic.

Software is that kind of mathematical mapping, from Input to Output, that is generally completely deterministic and mathematically ideal for function replicators.

The other easily replicated deterministic function is the Bourbaki Mathematician.

Nobody will miss the Bourbaki Mathematician. They had no original ideas.

Good mathematicians slave away on problems someone set for them.

Great ones invent new mathematics.

Nicholas Bourbaki is a fictional person for a very good reason.

Bourbaki has no creative soul so absolutely nobody will miss him.

Terence Tao thinks otherwise, which is okay.

Each to their own.

The bottom line is that replicable patterns are replicable.

Creation of the new pattern, the artwork not yet seen, is not replicable.

I grant that there will be a global difference of opinion on this matter. I happen to be one of the relatively few people right now who was born right-brain biased.

The right brain has zero difficulty generating new patterns.

Like anybody, I have two hemispheres and my left is considered smart. However, my right hemisphere, which is naturally mute, is orders of magnitude smarter.

I do have a daily struggle getting my left brain to understand simple stuff.

This used to be an enormous daily struggle but it is not any more.

Now I program the right question before I go to sleep.

Then I wake up with the answers and write them down. This cuts out a lot of arguing with myself during waking hours. I do not need to waste time with it now.

If such ideas shock you then I suggest you read this book.

Iain McGilchrist performed a great service in explaining how minds work.

What matters for our mundane stock picking exercise, is not the philosophy but a very simple operative principle with artificial intelligence:

Humans do the open world well and machines do closed worlds well.

Whether the world of discourse is open or closed can be decided operationally by observing which corporate strategy pivot is a winner in terms of survival.