The Future of Computing

Over the last hundred years we went from mechanical computers to punch cards and then to digital vacuum tubes. Eventually in 1947 the first transistor was invented at Bell Labs. With the advent of transistors, digital computers took off and ever since compute scaled exponentially. Nowadays, the most of computer chips are digital, where computing is based on electrons moving in silicon. It was not always working this way and for sure it won’t be the case in the future. Here are 3 types of computing which may eventually replace our traditional computers.

Light

The first technology, which offers a paradigm shift in computing, is Photonics, and if we think in scale - Silicon Photonics. The main idea is that you can perform computing using light (photons instead of electrons). Many people are heavily betting on this technology, mainly because Photonics allows to overcome several bottlenecks of the modern computers.

The key feature of light is that we can parallelize computing in the frequency domain - and this is huge for AI. That’s possible because light has an amazing property of having different colours that do not interfere. We can encode and process information at various colours (wavelength) simultaneously, which enables massive parallelism of computing. And that’s very fast… at the speed of light :). There are already many exciting photonics startups. I’ve summarised them in the table:

→The full list of top photonics startups to watch is available on my Patreon.

Radio Waves

When I first heard about the concept of computing with light, I was thinking why can’t we do the same with radio-frequency waves, those which we use for communication (e.g. 5G) - and in fact we can! There are already multiple publications describing this concept.

Simplifying the matter, both radio waves and light are basically electromagnetic waves. The main difference is in their frequency. Actually, we can apply similar to photonics principals for computing here. In terms of technology, RF is definitely cheaper than optics because it is widely used for communication. The disadvantage of RF compute is that there is less spectrum available - light has THz spectrum, while RF has only GHz spectrum. As light has much more bandwidth, this practically means that we can achieve higher degree of parallel computing with photonics compared to RF. While many people are betting on photonics chips and there are many commercial products available, RF compute is still in the research phase.

Biological

The 3rd way to compute is biological, actually putting a brain in a dish, which is quite controversial. The idea is to use live human neurons for computing power. It is actually the original idea from the Matrix: machines use humans for computing, and the Matrix is a part of the software interface to the brain. This was in the original script, but then it was decided that the average audience at that time won't understand it - and the idea was scrapped.

Now, turns out it’s possible! There is a startup in Australia which is actually growing live human neurons and integrating them with traditional computer chips. These efforts are done in order to advance AI. The startup is Cortical Labs and the chips they are building is so-called DishBrain. They grow groups of neurons (brain cells) on a silicon computer chip. The advantage of this approach for AI is that the neurons can think on its own, grow, learn, adapt, replicate or even die depending on the needs of the system. The biological and the electrical parts interface with each other by means of electrical pulses, similar to what happens in our brain.

The Cortical Labs first proof of concept is to train their chip to play the game of Pong. Neurons live in the game where they are constantly stimulated. When they act wrong - they receive random stimuli, when right - they get defined stimuli → this actually confirms the model according to Friston’s Free Energy Principle. If you are interested to learn more about it, here is the link to an amazing episode of Lex Fridman podcast.

→ The full version of this letter is available on my Patreon.