Uniting organic computers – our brains – with their electronic counterparts has moved farther from science fiction and closer to reality, thanks to recent bioengineering innovations.
The first is an artificial synapse, devised by researchers at Stanford University and Sandia National Laboratories.
Since synapses are mainly empty space, why is this a big deal? Because synapses are the places in your brain where thinking and memory happen.
A brain cell, or neuron, is an octopus-like assembly. Trailing away from the cell body are two kinds of filaments. Axons carry electrical impulses away from the neuron’s body to be sent to other neurons; dendrites receive electrical impulses from other neurons and direct them toward the neuron’s body. Axons and dendrites end in little pods and the pods face each other across tiny spaces called synapses.
Biochemical messages are sent from axons to dendrites across those synapses. But a synapse is more than empty space.
The axon’s side of a synapse is packed with tiny chambers holding chemicals called neurotransmitters. The dendrite’s pod surface is studded with sockets – “receptors” – where those neurochemicals nestle after they’ve crossed the synapse. An electrical impulse coming down the axon from the cell body tells the axon which and what amount of neurotransmitters to release. The chemicals cross the synapse and settle into the dendrite’s receptors. The kind and amount of the chemicals picked up by the receptors send a particular electrical signal back to the receiving neuron’s cell body.
When a particular event sets off an electrical pattern uniting a group of neurons through their synapses, a memory is formed. If the event happens again, reactivating that pattern, the brain needs less energy to process that signal – and the memory is strengthened. If you see Steve at work every day, it takes less mental energy to quickly recognize Steve when you see him at the post office or the ballpark.
A neuron’s gaggle of axon terminals, and its up to 200,000 dendrites, make it possible for a single neuron to be a part of thousands of these memory patterns. With an estimated 100 trillion synapses in your brain, it’s not surprising that you can remember everything from arithmetic tables to the names of the kids on your high-school baseball team.
HOW A COMPUTER MEMORIZES
Computer memories, though, are different – and their difference gives them limits. A computer remembers things as a series of ones and zeroes. These electronic digits have to be stored in places physically separate from the places where chips process them. So a computer’s memory is limited by the physical space designers give to it and the cost of manufacturing memory chips.
This is making computer engineers nervous.
In the past, they could rely on Moore’s Law, which holds that the number of transistors – a computer’s equivalent of neurons – on a chip will double every two years. But transistors’ size is now at the nanoscale and fast approaching the limits of manufacturing’s ability to make them smaller.
Moore’s Law is expiring just at a time when we’re expecting robotic personal assistants and smart cars to manage our lives and take on more and more complex tasks, all of which require prodigious amounts of digital memory. As Technology Review magazine asked in a 2016 article, “Moore’s Law is dead. Now what?”
The Stanford-Sandia team looked for the answer inside our heads.
Using cheap, organic materials – mostly carbon and hydrogen – the group made what it calls an electrochemical neuromorphic organic device, or ENODe.
The device’s mechanism emulates a synapse but resembles a battery. The ENODe has battery-like terminals, and protons flow between them through a salty fluid. The terminals are analogous to axons and dendrites, and chemical ions traveling through the fluid take the place of neurotransmitters.
Perhaps most important, the team was able to send 500 distinctly separate electrical pulses through its device. This translates to 500 different meanings of the electrical signal, all of which can be stored separately in one of these artificial neurons. In contrast, a computer’s transistor has only two electrical states; it’s either on or off. Also, early versions of the ENODe “remembered” the specific degrees of charge for more than 24 hours – longer than some of us can remember to pick up milk on the way home from work.
It did that not only combining memory and processing in the same physical space, but also using only about a tenth of the electricity that a conventional computer does.
When researchers simulated the ability of a theoretical array of ENODes to learn, they found that the array would easily master the task of recognizing handwritten digits from zero through nine with more than 95 percent accuracy.
Already, the researchers can make ENODEs about five-millionths of an inch in size. That’s far bigger than neurons or even today’s transistors, but still small enough to pack an array comprising a functional memory into a small space. The device’s ability to store hundreds of bits of information at the same time could be critical in designing computers that can deal instantly with the floods of visual and audio information that we expect our electronic assistants and self-driving cars to handle with ease.
But could the ENODe also be an interface between brain and computer?
It is promising.
The ENODe is made of materials that mimic human tissue’s flexibility and squishiness. Just as important, it operates on very low electrical voltages. In contrast, an artificial synapse developed this year by a European group needs almost a volt of electricity to store information – perilously close to the 1.5 volts that break down body fluids in which the brain is bathed.
PIPELINE TO THE BRAIN
Massachusetts Institute of Technology researchers have taken a simpler route to getting information out of, and into, the brain.
The group has developed a fiber no wider than a hair that can send electrical signals, light pulses and fluids into the brain and bring the same back out. At the fiber’s core, an optical waveguide, similar to a fiber-optic cable, is paralleled by two fluid channels and six electrical threads.
The group developed the device as a new tool to learn about brain functions and regions. By sending light into various parts of the brain, neurons there are stimulated. The electrical pathways they fire, and mental or physiological functions they engage, can tell researchers new things about what different parts of the brain do and how various parts interconnect.
But the implications are much broader. Being able to pump fluids through these tiny tubes could channel anti-cancer drugs directly into the center of tumors or make it possible to drain an aneurysm without brain surgery. The power to send electrical signals into the brain could eventually bestow the power to control movement or even thoughts and emotions – or, in a distant future, be able to send our own thoughts or emotions to a computer or another person.
The ability to insert a tube into the brain isn’t new. What makes MIT’s device different, and better, is that it’s not only as thin as a hair, but it’s also as soft and flexible as brain tissue itself.
To reach that goal, the scientists needed something soft that conducted electricity. They engineered a polyethylene conductor, doped the material with graphite flakes, and then compressed it in layers like the thin layers of dough in a French pastry.
As a result, inserting the tiny tube is likely to do far less, if any, collateral damage as it travels to its target than today’s stiff wires would.
The fibers could be left in place for a period of time to deliver repeated doses of a medicine or observe how brain regions behave under different conditions over time. Also, the fibers are so small and unobtrusive that several could be inserted into a brain at the same time to study how various regions interact under different conditions or events.
To test their invention, the MIT team threaded the fibers into the brains of mice. They pumped a chemical into the brain that sensitizes neurons to light. Then the researchers shot light pulses up the optical cable. The sensitized neurons created patterns of brain activity that the scientists then recorded using the electrical conduits in the fiber.
The fibers are so small and unobtrusive that several could be inserted into a brain at the same time to study how various regions interact under different conditions or events.
The research team’s next challenge is to create materials even softer and more flexible to enable even more detailed brain research. But the unspoken implication of their work remains the possibility of even more intuitive and sophisticated two-way electrical, chemical and optical communication with a working brain in real time.
TALKING BRAIN TO BRAIN
MIT’s approach is still a little low-tech for Elon Musk.
Having tackled electric transport, home power storage and the commercialization of outer space, he’s ready to confront the complexities of the human mind.
Musk and a few partners have formed Neuralink, a venture that will try to figure out how to implant what it calls a “neural lace” into human brains to enable brain-to-brain communication. Musk says that people can receive his implants by 2030.
As Musk sees it, transferring ideas by talking or typing is literally 1 million times slower than the brain’s speed in processing information. Time’s a-wasting. By embedding a network of devices in our brains a fraction as big as a grain of sand, Elon thinks we could transmit thoughts directly – either between people or, eventually, with an artificial intelligence. Elon says his goal is what he calls consensual telepathy.
Musk’s notion parallels a recent project announced by Facebook: Within two years, Facebook has said, it will have a working prototype of a skullcap studded with electrodes to gather your thoughts and automatically print them on paper.
Given the current crude state of reading brain signals from the outside, neuroscientists have scoffed at Facebook’s deadline. They also have scoffed at Musk’s 10-year schedule – but not at his idea.
Still, there are caveats.
Finding medical facilities willing to do brain implants in healthy people for the sake of experiment will be difficult at best. Experts have suggested that Musk find a medical need and pursue his idea through that venue instead.
For example, imagine someone with paralyzed legs who wants to go to the kitchen for a snack. With Musk’s Neuralink, the person could transmit the intention to go to the kitchen to an artificial intelligence, which then would operate a set of electrometrical “pants” that would walk the person to the kitchen without the person having to consciously control each movement of each leg and foot. The human and electronic minds would fuse to achieve a goal.
Currently, researchers are discovering the neurobiology of intention. So, it’s possible that Musk’s Neuralink would someday be able to distinguish between “I want to go to the bathroom” and “I want to go to the kitchen and make a peanut butter sandwich.” The assistive technology could take it from there.
The mind meld that produces that stroll to the kitchen is years, and possibly decades, away. But these innovations have brought it closer. TJ