The mere idea of moving a hand or foot triggers measurable activity in the brain. People with physical disabilities could therefore communicate or move again thanks to brain-computer interfaces. Elon Musk has now announced that his company Neuralink has inserted a computer chip into a human’s brain for the first time. Surjo Soekadar, Professor of Clinical Neurotechnology at the Charité in Berlin, considers the technology to be promising – but he also sees risks.
SZ: Elon Musk announced that his company Neuralink had inserted a computer chip into a human’s brain for the first time. Did that surprise you?
Surjo Soekadar: No, not at all. Neuralink received approval from the US authorities for this last year. And the company has been working on this project for several years with more than $500 million in funding. However, the approval came surprisingly quickly for me.
Why?
Neuralink is working on a completely new system, with components that have never before been used in this form for the implantation of brain-computer interfaces. A wireless chip in the brain with a lithium battery. A fully automated robot that implants this chip. Elon Musk is already thinking about industrialization, i.e. large-scale use. This has never happened before. But brain-computer interfaces themselves are nothing new. Research into this has been going on for 50 years.
In the Neuralink study, paraplegic people were asked to control a computer or smartphone with their thoughts. Is this realistic?
Yes, that is completely realistic. Research has long demonstrated this. With the help of brain-computer interfaces, people could control devices and computers in their environment – and thus communicate again. Other test subjects were able to control exoskeletons or prostheses on their bodies. People who previously couldn’t walk were able to move again. This research could significantly improve the quality of life for people with nervous system injuries. But as I said: This is just research and is not used on a long-term basis with patients.
Surjo R. Soekadar is Professor of Clinical Neurotechnology at the Charité in Berlin and researches brain-computer interfaces. He and his team were able to enable paraplegics to eat and drink independently again using a brain-controlled exoskeleton on their hands. Soekadar focuses on methods that do not require opening the skull.
(Photo: Pablo Castagnola/ Einstein Foundation Berlin)
Neuralink’s brain chip has 1,024 electrodes attached to it that are connected to nerve cells in the brain. How does this work?
First of all, you have to say: 1024 electrodes is a lot. The systems currently available use significantly less. These electrodes hang on threads. And the robot implants the electrodes in a relatively small area in the outer layers of the cerebral cortex. The chip can make contact with individual nerve cells via the electrode contacts and record their activity. The researchers at Neuralink can then wirelessly control the chip from outside.
And how can a person then walk again or control devices using thoughts?
There are two methods. Neuralink combines both. In the first method, the brain learns. Take, for example, a paraplegic person. The goal is for this person to be able to walk again. She will then receive a prosthesis or exoskeleton on her legs. The person then has to imagine themselves moving their legs. The chip records the corresponding signals from the brain. The exoskeleton is then moved as if the patient were walking. The desired behavior is rewarded. The brain then learns which signals it needs to send in order for this running movement to be initiated. The patient gradually learns to modify brain activity to improve control of the computer system.
And the second method?
In the second method, the computer learns, more precisely an artificial intelligence (AI). This method also requires the person to imagine themselves moving their legs. The chip records the patterns that arise in that person’s brain. This information can then be used to train an AI. Once this is done, the chip can transmit the appropriate signals to an exoskeleton that carries out the desired movements. This means that it is not the brain that learns, but the machine that learns. The combination of these two learning methods is particularly promising. By the way, this can also be done without an exoskeleton. The chip in the brain can also forward the signals to another chip in the spinal cord. This chip stimulates the corresponding patterns there. In this way, the spinal cord injury of a paraplegic can be bridged. Researchers from Switzerland demonstrated this last year. Neuralink is also planning this.
Every brain is different. Isn’t that a problem for the widespread use of this technology?
Yes, of course, that is a problem. But the AI technologies we now have allow us a kind of personalization. But every AI model must be adapted and trained to an individual. If you have the appropriate capacity, that’s not a problem.
How do you know that the robot hits the right nerve cell in the brain?
You don’t even know that. The robot just has to avoid the blood vessels in the brain. It implants where it causes as little damage as possible. There are 86 billion neurons in the brain. They are all connected to other nerve cells with around 10,000 synapses. So there is a lot of redundancy. So if you implant the electrodes in the motor area of the brain, then you will hit a nerve cell that has to do with what you actually intend. The algorithms then have to be trained accordingly. It is still unclear whether the Neuralink chip’s more than a thousand electrodes can achieve so much more. Neuralink’s data, which they have from experiments with pigs and monkeys, is not public. So we don’t know how much information Neuralink can read from the brain.
What else is unclear?
It is completely unclear how the test subject will tolerate this implant. Rejection reactions could occur. Or the signal quality decreases over time. The crucial question, however, is how much these implants actually improve the quality of life of patients. So this study goes beyond experiments and can the chip really be used in patients’ everyday lives? All of this is unclear. The psychological dimensions have not been researched either.
Can you give an example?
In a previous study unrelated to Neuralink, some patients had to have their implants removed because of infections. These people reported that it felt like a second accident. Through the brain-computer interface they were able to move again. And suddenly they had lost that ability a second time. I also see the danger of psychological dependence on this technology.
What do you mean by that?
Basically, we are talking about patients who need special protection. People who are desperate because of their illnesses, who quickly say: “I’ll go along with everything.” Several thousand people applied to take part in the Neuralink study. You have to be careful not to take advantage of these people. The technology also makes other applications possible, for example controlling the smartphone with the power of thought. And if you think about it further, people with a chip in their brain will eventually be able to think better and move better. This will change people’s self-image and personality; they will become psychologically dependent on technology. And of course also from a company like Neuralink.
Elon Musk’s goal is for people to be able to enter into a symbiosis with AI. Is this some nonsense from Silicon Valley or does it really have medical potential?
Brain-computer interfaces have one hundred percent medical potential. This is also the path Elon Musk is on and the reason why he is allowed to do all of this. But he ultimately wants to expand this beyond medical applications. Hence the surgical robot. Neuralink is a commercial company. And I can very well imagine that as this technology becomes more widespread, the inhibition threshold for using such systems for non-medical purposes will also decrease. Although we are talking about a time horizon of ten to twenty years, we have to be prepared for that. We should ensure that these chips are initially reserved for medical purposes, for example through prescription requirements and access restrictions, until we really understand what the social impact could be.
When will we next hear from Neuralink?
How quickly the test subject learns to use the chip depends on how often they are trained and how old and fit the person is. This will take a few weeks. The training of the AI too. I expect that we will see the first demonstrations in April or May. Overall, however, it may take another five to ten years before these types of human-computer interfaces become available as approved medical products in hospitals.