Quantum Computation vs Brain Inspired AI

At work the other day I was having a conversation with a couple of my fellow teachers discussing the future of artificial intelligence (AI). One of the teachers was arguing that to develop true, generally intelligent AI we would need to use quantum computers. They argued that the use of quantum computers would in fact allow us to develop AI more efficiently with less energy waste. I, in contrast, argued that the necessity to bring the quantum computing chips down to almost absolute zero to get them to work would in fact be less environmentally friendly and less cost effective, although the quantum computer would perform the calculations expeditiously. Although I still believe that, Google AI Gemini, when asked, responded that using quantum computation would in fact be beneficial to generating more energy efficient AI. As far as who is correct, only time will tell.

Brain Inspired AI

I believe, instead of using quantum computing, we need to create brain inspired AI algorithms to bring the energy cost down. If we could make an AI algorithm that better models theories of how the brain functions, we could create more energy efficient AI models. The human brain is surprisingly efficient, needing less training time than the typical AI algorithm to learn about the environment. The National Institute of Standards and Technology (NIST) wrote, “The human brain is an amazingly energy-efficient device. In computing terms, it can perform the equivalent of an exaflop — a billion-billion (1 followed by 18 zeros) mathematical operations per second — with just 20 watts of power.” (https://www.nist.gov/blogs/taking-measure/brain-inspired-computing-can-help-us-create-faster-more-energy-efficient) This is an incredibly efficient computation device, and you are blessed to have one nestled within your skull, no laptop required. If we could create an accurate theory of how the brain works, we could model that theory in silico, on a computer, and thus create an extremely efficient artificial general intelligence (AGI).

The problem is that we don’t yet have an accurate model of our brain, although we are trying. In fact, the artificial neural networks (ANNs) commonly used for our AI algorithms today were an attempt to model our brain’s calculations in silico. The ANNs are an extension of a model called a perceptron. The perceptron was developed based on what we knew about the visual cortex of our brain. The details of the perceptron will not be discussed here, but if you want to read the original paper detailing its creation here is its link: https://www.ling.upenn.edu/courses/cogs501/Rosenblatt1958.pdf. We have learned a lot more about how the brain functions since the 1950s, but we have not developed many new theories that can be translated for use in computation. I want to see more theoretical neuroscientists working on creating new theories about how the brain works, but that is not to say we have not made any progress on this path. There is one company working on exactly this, creating brain inspired AI, Numenta. You can read more about them on their website, https://www.numenta.com/, but essentially, they have created a theory of the brain called the Thousand Brains Theory. They wrote about this theory in a book called A Thousand Brains: A New Theory of Intelligence (https://www.amazon.com/Thousand-Brains-New-Theory-Intelligence/dp/1541675819) which is where I learned about this company and their theory.

Numenta has not only created this theory of the brain but has done a lot of work to create better AI using the theory. Their AI algorithm has a very different architecture from the ANN, and I still don’t know exactly how to use it in Python, but they say that it can be trained efficiently using a CPU. The fact that it can use a CPU, rather than the commonly used GPU necessary for our current deep-learning ANNs, is a great leap forward in AI and energy efficiency. I want to see more progress toward getting their AI to work, because if it works, it would be evidence that maybe their theory of the brain is correct. There is evidence that it does have its flaws, however, but it is a step in the right direction, and with more competition from other companies we may one day understand how our brain works.

Quantum Computers

The use of quantum computers to create more efficient AI has some merit as well. I can’t say I fully understand, and don’t believe anyone understands how a quantum computer works because it deals with quantum theory, a theory that makes no sense in this macroscopic world we live in but will do my best to explain how it can, in theory, help create better AI.  Disclaimer: I am not a theoretical or quantum physicist, so do your own research on this and all the topics I have written about. Question and then research everything!

The advantage of using quantum computers is that they can perform the computations in a fraction of the time of a normal computer. Normal computers use 1s and 0s, bits, to perform their computation. They use these bits to compute everything from math to displaying images on a screen. The bit can either be a 1, on, or a 0, off. If you want to learn more about how computers use bits to compute, check out this Khan Academy video: https://www.khanacademy.org/computing/computers-and-internet/xcae6f4a7ff015e7d:digital-information/xcae6f4a7ff015e7d:bits-and-bytes/v/khan-academy-and-codeorg-binary-data.

Quantum computers on the other hand use qubits, which are like bits, but they can be 1 or 0 or both 1 and 0 at the same time. This is due to the quantum mechanical principle called superposition. With superposition a quantum particle takes on all states at once, like a Schrodinger’s cat being both dead and alive, and only chose a state when measured. So, by being both a 1 and a 0, a quantum computer can zoom through all the computations necessary with way less bits/qubits needed by traditional computers. This speed of calculations will make training an ANN theoretically take second rather than potentially hours taken for some algorithms now. This would lead to a dramatic decrease in compute time and energy consumption.

There is another quantum principle called entanglement which can also contribute to a quantum computer’s decreased computation time called entanglement. When a quantum particle becomes entangled with another particle, even if the particles are a billion miles from each other, if one is measured as being a 1 the other instantly shows a 1 as well. This property can increase computation time by allowing qubits that need to have the same value do so immediately without any energy draw or heat being released.

How quantum mechanics works still baffles my mind, but these are some of the methods used by quantum computers to expedite calculations. One problem is that the chips that hold the qubits need to be kept at temperatures a few thousandths above absolute zero to keep them working appropriately. That is insanely cold, and to maintain that cold temperature requires a lot of energy. This means that a quantum computer may still have a big energy draw from the power grid, even if it can train an ANN faster.

There are multiple challenges quantum computers have to overcome to be useful in today’s society, but even if scientists overcome all the obstacles, quantum computers may never be able to be brought home for daily use. Because of the energy expenditure and the size of the quantum computer, they will have to be housed in government or corporate offices with a lot of money. Software developers, like me, may never have the ability to play around with quantum computation. This means if quantum computers are used to make AI, the AI will be controlled by the government or giant corporations which may lead to its own problems. Ultimately, I don’t think quantum computers are the way to produce better AI, or AGI. I think, if we are to create AGI, we should do so to better understand the brain and mind and not make money. We should focus on making brain inspired AI to test theories of how the brain works, not with the goal of taking away human held jobs.

What are your thoughts about this topic? Leave a comment down below and share this article with a friend. Also subscribe to read about more topics around the intersection of biology and computer science.