Artificial intelligence seems to be superior to humans in many areas, but one of the aspects in which the human mind is superior is energy consumption. To perform a complex arithmetic operation, the human brain needs 20 watts of power, while a supercomputer for a similar task requires nearly a million times as much power. This computer runs at 20 megawatts. Couldn’t it be done better, researchers have been wondering for a long time? Can’t we design devices that closely resemble the human biological brain?
One of them is Cristian Nigues, who does research at TU Twente on so-called molecular switches. His team recently had a breakthrough, that is Described in a trade journal natural materials. It’s time to take a look at the Nijhuis lab, as it works to design and measure new molecules.
Nijhuis calls himself an “architect at the molecular level”. It draws molecules and if they look interesting enough, chemists can make such a molecule. It’s called synthesis. “Now we’ve found a molecule that can mimic synapse function,” Nigues explains. The brain is made up of about a hundred billion neurons, with each neuron connected to thousands of other neurons. Synapses are the communicators: every signal in the brain passes through synapses. Of these, there are an estimated one hundred Two hundred trillionwhich is the number 1 that contains fourteen zeros.
For Nijhuis, the interesting thing about these clamps is not their unimaginable quantity, but their efficiency. This is why the Synapse is such a great source of inspiration when designing new computers. And there is a great need for this: “Processing ever-increasing data streams is getting out of hand. Data centers consume energy and will only get bigger.” Nijhuis mentions a well-known AI example: recognizing a cat in a photo. AI is very good at this these days, but it takes a bit enormous amounts of computing power to train the models and then make the prediction “Is this a cat?” to be executed.
Nijhuis: “Our brains do it much better. They only use energy when an information pulse passes through the synapse, which means they can process a lot of data at the same time. And not only that, they’re also flexible and dynamic: synapses act as switches, but the thing What’s special is they can change all the time.So you want a molecular machine that also has those dynamic properties.A computer that, like the human brain, makes new connections every time it learns something new, and it strengthens or weakens existing ones, just what’s needed.
Precise manufacturing and measurement process
This simulation has now been achieved for a single synapse. The first step was taken two years ago, with the arrival of the molecular on/off switch. And now there is a key that is also learning from the past. “It was cool to see that our molecules behaved as we had hoped,” says Nijhuis with fire.
This exhilarating moment was preceded by a meticulous fabrication and measurement process. A very thin layer exactly one molecule thick is applied to a gold plate, after which it is separated again. Nijhuis details microchannels, alloys, oxide monolayers, and stable electrodes. Munter: “That’s it, a very simple manufacturing method.” In the lab, a student displays the slippery monolayer, with electrodes on both sides. It uses a multimeter to prove that the molecular layer does indeed work.
The basement of the laboratory was set up as a vibration-free Faraday cage, where the behavior of molecules could be measured with special equipment. Nejoes hopes that a single molecule will open the doors to a whole family of new molecules and materials. and finally to an artificial neural network in which the artificial synapses no longer process binary information streams (with zeros and ones), but operate in a similar fashion.
“This is unprecedented.”
There is still a long way to go, Neghis admits. Johan Mentink, who is doing somewhat similar research at Radboud University, thinks so, too. Mentink, who was not involved in Nigues’s research, described TU Twente’s discovery as a scientific breakthrough: ‘They have succeeded in making synapses on a molecular scale. This is unprecedented.
Maintink himself is working on, among other things, magnetic optical gratings, which switch based on light. The big advantage of this is that it is more economical. An important disadvantage is that it is not organic and therefore less suitable for medical applications.
And it is precisely in this latter direction that Nijhuis thinks with his organic invention, as well as all applications for which an abundant amount of energy is not available, such as self-driving cars or drones. Wouldn’t it be great if the body could use a super-powerful and economical computer made of soft materials instead of hard silicon wafers, Nijhuis dreams out loud.
For example, he mentions an implant that is placed on the visual cortex and can process information coming from the eye, as an aid for the blind. Careful experimentation with implants is already underway, but these chips are a rough solid with limited results (in this specific example): a person wearing a blind spot sees an image of a few dozen pixels.
Negues hopes that other useful, but less exciting, applications will soon come in handy, such as devices that can administer insulin based on a diabetic patient’s eating pattern and behavior.
Don’t run straight to the health store just yet, Nijhuis warns: “It will take years and years of basic research before we are ready for practical applications.”
