How to use quantum computing in AI and neural networks

Google’s announcement of the OpenFermion quantum chemistry package is one of the most exciting things in computing that I’ve heard of. It really makes me emotional that people have done this work and are sharing it with scientists.

Credit
Announcing OpenFermion

Ask yourself: how are new drugs (or other materials) discovered? Well, presumably, a lot of chemists do a lot of lab experiments, try out different molecules and combinations, test these, look at their effect under a microscope to figure out what’s going on, etc. (The Drug Discovery Process)

Now, (without quantum computers,) imagine building “classical ordinary” neural networks to replace the chemist. Well, your neural network would still need hands, eyes, and measurement tools - i.e. an automated “laboratory” device - to be attached to the neural network, so the neural network can perform these experiments.

That’s where quantum computers and the OpenFermion package from Google come in. One of the most important applications of quantum computing is simulating the world of chemistry, so that you can compute what molecules do rather than go into the lab and directly test it.

So, yes, quantum computing would be very useful and exciting in combination with neural networks.

Unfortunately, my knowledge of quantum computing is limited. For example, if we had extremely powerful quantum computers working (or perhaps even not so powerful ones) would we still need classical computers at all? A very interesting question that I can’t answer now.