Article written by Tijn van der Zant Ph.D., CEO Datamaister

Quite often, when someone asks me what I do, I say that I make computers smarter. Then I give some examples, like a talking computer assistant, an autonomous car, or an elderly care robot. When someone wants to know more, I have tried to explain the history of Artificial Intelligence (AI) and where we are now. Still, most of the time, I lose people when I talk about Alan Turing in the ’40 s or the Dartmouth Summer Research Project on Artificial Intelligence in 1956.

On the surface, it seems to have little connection with what is often heard in the media, although on a deeper level, it does… So, I started explaining the buzzwords from the press, and that does connect with a lot of folks.

In a series of articles, I explain the differences and similarities of some of the buzzwords. In the previous article, I discussed Data Science, Machine Learning, AI, and Deep Learning. In this article, I will discuss Quantum AI and Quantum Machine Learning.

Quantum AI & Quantum Machine Learning

Quantum AI, Quantum Machine Learning, and the upcoming field of Quantum Deep Learning are similar to the technologies described above. Except the algorithms run on a Quantum Computer. And that makes it a different beast.

Quantum Computers of sufficient size can make calculations that would take forever on regular machines. This opens up some incredible possibilities by itself but possibly even more so if we start combining this with the power of AI and Machine Learning.

A regular computer uses bits, a 0 or 1, to store data and to compute. A quantum computer uses q-bits, which can be 0 and 1 at the same time. If we have a problem of 3 bits, a regular computer has to go through all the combinations of zeros and ones, generating the following: 000, 001, 010, 011, 100, 101, 110, 111. This is 23 bits. For any problem having n bits, the classical computer must generate and test 2n solutions to find the best result. If the problem is, say 50 bits, that is 1.125.899.906.842.624 combinations. Since q-bits are in a superposition of 0 and 1 at the same time, it can ‘see’ all combinations of zeros and ones at the same time. Even if a classical computer can generate and test 1 billion combinations per second, which it usually cannot, it takes about 13 days to come up with an answer. A quantum computer can do that in seconds, or even less than a second. If the classical computer can test 1 million solutions per second, it already takes about 35 years to come up with an answer.

So why are we not Quantum Computing then all the time? That is because they are not very mature. There is a lot of noise in the computer, resulting in procedures when the computation must be performed many times to get an approximate answer. Quantum Mechanics is notoriously tricky to understand, and the programming of Quantum Computers is just as problematic. The last issue is that they are still exceedingly small, meaning less than 100 q-bits.

Quantum for solving specific problems over generic ones

My prediction is that in the 20’s we will see the rise of (online) quantum computers to solve particular problems but not yet generic ones. At the moment, large IT companies (IBM, Microsoft, Google, Alibaba, etc.) and research laboratories worldwide are working on creating a Quantum Computer that demonstrates Quantum Supremacy. This means that a Quantum Computer can compute something in a reasonable time that a classical computer cannot. Although Google claims they have achieved this, not everybody is convinced.

There are several ways we could combine Quantum computers and ML:

  1. Run ML algorithms on a quantum computer
  2. Use ML on a conventional computer to write quantum programs to solve specific tasks
  3. Use quantum computers to find the optimal settings and architecture for an algorithm that runs on a conventional computer.

As mentioned above, the first quantum computers will not be general problem-solvers, so not all types of algorithms can run on it. Possible candidates are path planning algorithms which might allow us to solve tasks such as optimizing logistics.

Quantum versus conventional computers

Quantum computers are not easy to understand; writing code for them is also different than usual. Using ML algorithms, such as genetic programming, which runs on a conventional computer, might be an excellent way to write programs for quantum computers.

One of the challenging tasks within conventional AI and ML is to find the optimal settings and architectures for specific algorithms, particularly with deep learning. As Quantum Computers are great at quickly finding optimal settings, potentially, they might be used here too.

In all these cases, much research still needs to be done, and once quantum computers become (more) available, we should see more development here too. It certainly is fascinating and something to keep an eye on.

 

Tijn van der Zant
tijn@dataMaister.com