Monday, November 24, 2014

How to build a quantum brain more conscious and powerful than a human brain

To begin with -- why analog quantum computing is more powerful than digital quantum computing,
AND what we need to do to actually build it.

Last month, I was invited to give a keynote talk at the Pacific Rim AI conference (PRICAI) on how to combine intelligence and quantum computing. Thanks Eli! The Pacific Rim is now moving 'way ahead of the US in many aspects of machine learning -- and Queensland, Australia, the site of this conference, is an important world center in quantum computing.

In summary: to build the most powerful possible intelligent system, one combines brain-like learning WITH the inherent computing power of the most powerful possible quantum computing.

Long ago, we learned that we get more power and more potential for learning by moving from a digital way of thinking to an analog way of thinking, form Turing machines to neural networks.
The first step in building quantum intelligence is to move form digital quantum computing to analog quantum computing.

For the conference, I talk about this for 40 minutes. The audio file (a mov file) and the 14 slides which go with it (in pdf) are posted on the web:

The key idea for analog quantum computing is very simple, in the end. Instead of using a spin
to represent a bit, |0> or |1>, use it to represent a continuous variable, |theta>.  That's just like modern neural networks, which get past 0 and 1, and make use of bounded continuous variables.

But.. we can't design computers until we can design, understand and model the basic logic elements they are made of. For classical computers, these are transistors  For quantum computers computing with spin instead of voltage, these are "spin gates." For analog quantum computing, these are TUNABLE spin gates -- basically just glorified polarizers. We can't understand how a big network of spin gates will work, when entangled particles flow through them.. unless we can at least predict how three or four entangled photons will behave, when they enter a simple network.

It turns out -- for all the talk about quantum computers with hundreds of qubits, there are only three groups in the world which have actually entangled three or more photons, with general spins theta.
No one on earth has actually done the crucial experiments with three photons which make it clear which model we should use in designing larger systems. the first key experiment is now underway
cited in my papers). And so... to get to analog quantum computing, the first necessary baby step is to really nail  down what happens in the full triphoton experiment I talk about at the conference.
And then... Shih has a new way to produce a hundred entangled photons; the pathway is open, if people are ready to take the next step.


My talk also briefly mentions how we can get even further than analog quantum computing, but maybe for folks in teh mainstream.. the challenge at present is to nail down analog quantum computing. One revolutions at a time...

No comments:

Post a Comment