Tech News: Intelligent material makes quantum brains a possibility

Since the 1940s computer scientists have tried to build more intelligent computers by simulating the large array of neurons in the neocortex of the human brain in artificial neural networks. Photo: Manjunath Kiran/AFP/Getty Images

Since the 1940s computer scientists have tried to build more intelligent computers by simulating the large array of neurons in the neocortex of the human brain in artificial neural networks. Photo: Manjunath Kiran/AFP/Getty Images

Published Feb 16, 2021

Share

SINCE the 1940s computer scientists have tried to build more intelligent computers by simulating the large array of neurons in the neocortex of the human brain in artificial neural networks (a series of algorithms that mimic the operations of a human brain to recognise relationships between vast amounts of data).

In combination with the ever-increasing computing power of computers, neural networks and autonomous deep learning (part of machine learning based on learning data representations), machines are now able to independently recognise objects and translate speech in real time.

The Boltzmann machine

Dell embarked on their first ever neuroscience “Brain on Tech” study this year

In the 1970s neural networks could only simulate a limited number of neural networks and were, therefore, unable to recognise complex patterns. In 1983 Geoffrey Hinton and Terry Sejnowski invented an unsupervised deep learning model, named the Boltzmann machine, based on the Boltzmann Distribution (a probability measure).

A Boltzmann machine is an artificial neural network of symmetrically connected, neuron-like units consisting of stochastically (non-deterministic) firing neurons that can learn complex probability distributions by adapting the synaptic interactions between neurons. Boltzmann machines represent a generic class of stochastic neural networks that can be used for data clustering, generative modelling and deep learning.

In artificial intelligence (AI), the computer needs to recognise patterns and learn new ones through machine learning software and neural networks. Unfortunately, a key drawback of software-based stochastic neural networks is the required Monte Carlo (repeated random) sampling, which scales incredibly with the number of neurons, thus increasing the computational time scales exponentially.

Consequently, Monte Carlo sampling is computationally expensive since it requires powerful, energy-hungry computer systems. It is thus important for the further development of AI that we find new strategies and approaches to store and process information in an energy-efficient way.

Neuroscience

Neuroscience does not really understand what consciousness is or how it works. Neither do scientists fully understand quantum mechanics. Could this perhaps be more than a mere coincidence?

It is, therefore, no wonder that the enigma of consciousness has led some researchers to propose that quantum physics could possibly explain it – a view that has been met with scepticism since it does not make sense to explain one mystery with another. But despite the scepticism, the mind has forced its way into early quantum theory.

One of the mysteries of quantum mechanics is that the outcome of a quantum experiment can change depending on whether or not we choose to measure some property of the particles involved. This created a dilemma: If the way the world behaves depends on how – or if – we look at it, what can “objectivity” and "reality" really mean?

Some researchers, therefore, concluded that objectivity is an illusion and that consciousness has to play an active role in quantum theory. Some physicists suggest that whether or not consciousness influences quantum mechanics, it might in fact arise because of it. Quantum theory is thus needed to fully understand how the brain works. Just as quantum objects can apparently be in two places at once, so can a quantum brain hold onto two mutually exclusive ideas at the same time.

In the past few years quantum effects have been connected to photosynthesis, a process fundamental to life on Earth. They apparently also play a role in other biological processes such as avian migration and olfaction. The tunnelling hypothesis developed in the context of olfaction has been applied to the action of neurotransmitters. It has also been proposed that general anaesthetic, which “switches off” consciousness, does this through quantum means, measured by changes in electron spin. A more recent theory outlines how quantum entanglement between phosphorus nuclei might influence the firing of neurons.

Based on the above progress in research, an important hypothesis in neuroscience postulates that classical mechanics cannot explain human consciousness, since intelligent materials such as our brains learn by physically changing themselves, and therefore are better explained through quantum mechanical phenomena such as entanglement (the quantum state of each particle of a pair or group cannot be described independently of the state of the others, including when the particles are separated by a large distance) and superposition (the ability of a quantum system to be in multiple states at the same time until it is measured).

Whether quantum physics lie beneath the functioning of the human brain and contribute to neural processing or not, the world of AI and computing will be changed for ever if scientists can construct materials and machines that can truly mimic the complex behaviour of the human brain.

The quantum brain

This is exactly why Alexander Khajetoorians, professor of scanning probe microscopy at Radboud University in The Netherlands, suggests that building a “quantum brain” based on the quantum properties of materials could be the basis for a future solution for applications in AI.

In Nature Nanotechnology of February 1, 2021, a group of scientists at Radboud University published a very significant paper under the title “An atomic Boltzmann machine capable of self-adaptation” in which they describe their investigation into whether a piece of hardware could do the same as neural networks, without the need of software.

Radboud physicists found that by patterning and interconnecting a network of single cobalt atoms on black phosphorus they were able to build an intelligent material that processes and stores information (learn) by physically adapting itself depending on the input, similar to the autonomous behaviour of neurons and synapses in a human brain. This remarkable discovery of the so-called “quantum brain” could be the foundation of a completely new generation of computers.

Self-adapting atoms

The discovery builds on the research done by Khajetoorians and colleagues in 2018 when they succeeded in storing information by manipulating the quantum state of a single cobalt atom (“An orbitally derived single-atom magnetic memory”, Nature Communications 9, September 25, 2018).

By applying a voltage to the atom, they could place the atom in a state of superposition (induce “firing”), where the atom assumes two states (the value of 0 and 1) simultaneously, much like a neuron, thus mimicking the behaviour of a brain-like model used in artificial intelligence. The atoms could thus be used to store information.

In addition to observing the behaviour of spiking neurons, they observed that the synapses changed their behaviour depending on what input they “saw”. The material learnt by itself by adapting its reaction according to the external stimuli that it received.

The functioning of the quantum brain

The discovery by the Radboud scientists introduced the possibility of a totally new way of processing and storing information by using an adaptive material.

The input entails that the quantum brain is constructed of a material with groups of linked atoms that are in a certain state (0 or 1). This input is then processed by the self-adaptive quantum brain through the changing of the state and link of the atoms depending on what input they “see” (e.g. the image of a human person). The material recognises the input and gives as output a human person, without the use of a computer program.

The future of the quantum brain

The Radboud researchers plan to build a much larger network of atoms, as well as use new quantum materials with the ultimate aim of eventually building energy-efficient, self-learning computing devices and machines. But to achieve this, they still have to solve the mystery of why the atom network behaves as it does. Of importance is that the first steps towards a quantum brain have been taken.

Although a sceptical Albert Einstein once rejected quantum theory in its original formulation with the now famous words: “God does not play dice with the universe,” the quantum debate has changed dramatically.

Michael J Biercuk, the founder and chief executive of Q-CRTL in Sydney, Australia, and professor of quantum physics and quantum technologies at the University of Sydney, quite correctly stated that “Quantum technology will be as transformational in the 21st century as harnessing electricity was in the 19th.”

Professor Louis CH Fourie is a technology strategist

*The views expressed here are not necessarily those of IOL or of title site

BUSINESS REPORT

Related Topics: