Blueprint for an artificial brain: Physicist takes nature as his model

Blueprint for an artificial brain
A nanocomponent that is capable of learning: The Bielefeld memristor built into a chip is 600 times thinner than a human hair. Credit: Bielefeld University

Scientists have long been dreaming about building a computer that would work like a brain. This is because a brain is far more energy-saving than a computer, it can learn by itself, and it doesn't need any programming. Privatdozent [senior lecturer] Dr. Andy Thomas from Bielefeld University's Faculty of Physics is experimenting with memristors – electronic microcomponents that imitate natural nerves. Thomas and his colleagues proved that they could do this a year ago. They constructed a memristor that is capable of learning. Andy Thomas is now using his memristors as key components in a blueprint for an artificial brain. He will be presenting his results at the beginning of March in the print edition of the Journal of Physics D: Applied Physics.

Memristors are made of fine nanolayers and can be used to connect electric circuits. For several years now, the memristor has been considered to be the electronic equivalent of the synapse. Synapses are, so to speak, the bridges across which (neurons) contact each other. Their connections increase in strength the more often they are used. Usually, one cell is connected to other nerve cells across thousands of synapses.

Like synapses, memristors learn from earlier impulses. In their case, these are that (as yet) do not come from nerve cells but from the electric circuits to which they are connected. The amount of current a memristor allows to pass depends on how strong the current was that flowed through it in the past and how long it was exposed to it.

Andy Thomas explains that because of their similarity to synapses, memristors are particularly suitable for building an artificial brain – a new generation of computers. 'They allow us to construct extremely energy-efficient and robust processors that are able to learn by themselves.' Based on his own experiments and research findings from biology and physics, his article is the first to summarize which principles taken from nature need to be transferred to technological systems if such a neuromorphic (nerve like) computer is to function. Such principles are that memristors, just like synapses, have to 'note' earlier impulses, and that neurons react to an impulse only when it passes a certain threshold.

Thanks to these properties, can be used to reconstruct the brain process responsible for learning, says Andy Thomas. He takes the classic psychological experiment with Pavlov's dog as an example. The experiment shows how you can link the natural reaction to a stimulus that elicits a reflex response with what is initially a neutral stimulus – this is how learning takes place. If the dog sees food, it reacts by salivating. If the dog hears a bell ring every time it sees food, this neutral stimulus will become linked to the stimulus eliciting a reflex response. As a result, the dog will also salivate when it hears only the bell ringing and no food is in sight. The reason for this is that the nerve cells in the brain that transport the stimulus eliciting a reflex response have strong synaptic links with the nerve cells that trigger the reaction.

If the neutral bell-ringing stimulus is introduced at the same time as the food stimulus, the dog will learn. The control mechanism in the brain now assumes that the nerve cells transporting the neutral stimulus (bell ringing) are also responsible for the reaction – the link between the actually 'neutral' nerve cell and the 'salivation' nerve cell also becomes stronger. This link can be trained by repeatedly bringing together the stimulus eliciting a reflex response and the . 'You can also construct such a circuit with memristors – this is a first step towards a neuromorphic processor,' says Andy Thomas.

'This is all possible because a memristor can store information more precisely than the bits on which previous computer processors have been based,' says Thomas. Both a memristor and a bit work with electrical impulses. However, a bit does not allow any fine adjustment – it can only work with 'on' and 'off'. In contrast, a memristor can raise or lower its resistance continuously. 'This is how memristors deliver a basis for the gradual learning and forgetting of an artificial brain,' explains Thomas.


Explore further

Blood simple circuitry for cyborgs

More information: Andy Thomas, 'Memristor-based neural networks', Journal of Physics D: Applied Physics, dx.doi.org/10.1088/0022-3727/46/9/093001 , released online on 5 February 2013, published in print on 6 March 2013.
Citation: Blueprint for an artificial brain: Physicist takes nature as his model (2013, February 26) retrieved 15 June 2019 from https://phys.org/news/2013-02-blueprint-artificial-brain-physicist-nature.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.
0 shares

Feedback to editors

User comments

Feb 26, 2013
A first step, yes, but IMHO Pavlov's dog response cannot be compared to a human learning how to do calculus. After all, what did the dog learn? It learned to salivate when no salivation was called for! Not only does the capability for learning have to be engineered, but the ability to know what has been learned and to make deductions based on what has been learned. Genuine artificial cognition is still a ways off . . .

Feb 26, 2013
The Bielefeld memristor built into a chip is 600 times thinner than a human hair

This statement is meaningless. It might have value if I was trying to explain the concept of CMOS to my grandmother, but it has no place on a science-centric website. In the world of science, we do NOT measure size relative to human hairs - please use the metric system.

Feb 26, 2013
A first step, yes, but . . . cognition is still a ways off . . .


Agreed- learning is not cognition and cognition is not necessarily required to learn. The step taken here with memristor technology is to develop something that can remember a previous state and bring that information to bear in a future state. Both learning and cognition require this fundamental activity. A single, tiny electronic element that can do this is a breakthrough for both areas.

Feb 26, 2013
To most people calculus represents an understanding of mathematics that is almost beyond comprehension. People can understand training a dog. Hence the reference to Pavlov's dog.

As for human hair comparisons, while I agree the dimensions should have been in nm, the typical person doesn't have a good feeling for the unit. Virtually everyone understands the thickness of their hair.

Do we want a website only for the scientifically literate, or one that might get the attention of the less (scientifically) literate and help to communicate scientific breakthroughs to the masses? (JMO)

Feb 26, 2013
It's velocity sensitive. A bot would be able to sense how hard/soft it was hit with a certain stimulus over time rather simply sensing a stimulus as a singular event.

Feb 27, 2013
Any real system simulating neuron synapses have to account for not only the strengthening of connections from previous excitation's, but also the weakening of those connections from inhibitory connections. It is quite possible that the system described has that capability (by using current in the reverse direction for example), but the electronics seem awkward to me. Could someone here explain how inhibition would or could work?

Feb 27, 2013
Virtually everyone understands the thickness of their hair.


Yes, but only in relations that are small, like twice or half the thickness. 600 times smaller becomes just a small and meaningless number.

Like, what is 600 times longer than your thumb? Once you go beyond one order of magnitude in difference, the comparison loses meaning.

I think author's here should set a standard that EVERY measurement should be expressed properly in metric


The authors here mainly just ctrl-c ctrl-v the articles from other sources, or, they're not real authors at all but just crawler programs that do it automatically.

Feb 27, 2013
Glad someone is using memristors after Hynix (and HP) scuttled production of the tech that would have massively out performed flash memory.
http://www.tomsha...986.html

Feb 28, 2013
Low consumption, co-located memory and program, and asynchronicity are good to simulate biological memory, but hardly enough to emulate the brain. The brain is a self-organizing dynamical system. It works because optimally short dendrite trees produce least-action dynamical trajectories, which is where self-organized invariants are found. This is a thermodynamic process, not a circuit or algorithm. Short dendritic trees were theoretically predicted in 2011 and independently confirmed in 2012 with a large experimental base. Not clear how memristor memories are supposed to produce least-action and self-organization.

Mar 02, 2013
As for human hair comparisons, while I agree the dimensions should have been in nm, the typical person doesn't have a good feeling for the unit. Virtually everyone understands the thickness of their hair.

Would that be an Mediterranean human hair or a finer Norwegian human hair? Hmmmm - nm would be better...

Mar 02, 2013
...Do we want a website only for the scientifically literate, or one that might get the attention of the less (scientifically) literate and help to communicate scientific breakthroughs to the masses? (JMO)

THere are numerous science specific sites on the internet. My thought is that physorg is not intended for that sort of reader. Most articles I've read give reference to articles on one of these other "science" sites.
Therefore, those who comment on this site, all puffed up with scientific know how and disdainful of other, less "educated" questions/opinions are either poseurs or Sheldon Cooper.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more