Donald Clark Plan B: Musk. Neuralink's Brain-Machine interfaces and Starlink

Saturday, November 13, 2021

Musk. Neuralink's Brain-Machine interfaces and Starlink

Elon Musk has many strings to his bow but one is of particular interest to learning, started in 2016 - Neuralink. An array of thin fibres has been developed that can be inserted into the brain so that data can be read, analysed and turned into actions. Significant progress has been made and, with the release of the Pager monkey video in April 2020, where neural activity was decoded to play a computer game. This has opened up debate around invasive (and non-invasive) technology for learning. The idea of plugging the brain into learning experiences is one small step closer.

Brain-machine interface

Research in this area has been going on for over 100 years, since the 1920s. More recently brai implants have had great success in

In a video showing a nine year old primate, called Pager, playing a computer game. At first it plays the game with a joystick, rewarded by a banana milkshake through a tube. It learns the game and plays well. The joystick is then unplugged and the primate plays only through the fibre array of 2000 fibres, each about one twentieth the width of a human hair, inserted into its brain, in the region known to control hand and arm movements. Pager even continued to play the game with its mind ignoring the joystick. The intentional data from the fibres is interpreted by AI and it continues to play the game well. The efficacy of the array was obvious.

It’s initial use is targeted at paraplegics, quadriplegics, stroke victims and those with locked-in syndrome, who will be able to control their own movements and interact with the world. The control of wheelchairs and exoskeletons are obvious applications. Epilepsy and the monitoring and prevention of seizures is also being researched along with other neurological disorders such as deafness and blindness. They also have mental illness in mind, such as ADHD, addiction and depression.

Beyond this a general application is control of computers at the speed of thought. This wireless interface short-circuits the clumsy use of ‘meat’ fingers and thumbs. This causally links the brain directly with all online resources. The goal is to give people communication capabilities via text of speech synthesis.

Learning

Once the brain is linked to the internet all learning resources are available but with a frictionless interface. This is obviously useful for those with disabilities but it reduces the cognitive load of traditional interfaces for everyone. This in itself should make learning quicker. Brain to device to internet opens up the possibility of more direct access to learning opportunities not only by those with disorders but all learners.

More than just access, if a feedback loop is possible, personalised learning could be made available with immediate formative assessment and feedback through adaptive learning systems. AI not only reads what you do and know, it provides support and opportunities to learn how to overcome any problems you have in learning. The future development of such systems is now underway.

Starlink

Musk has also launched thousand of low-level satellites called Starlink (a division of SpaceX) to provide low-cost 5G internet at any spot on the planet. The plan is to launch 42.00 satellites by 2027. The possibility of delivering quality online learning to anyone, anytime, anyplace may be realised by this initiative. You need a satellite dish and terminal and these have already been shipped to a large number of paying customers.

Critique

There are those who think this approach has real limitations and that all talk of reading intentions beyond motor skills is misleading, never mind interpreting semantic thoughts through language. Some think that invasive technology is too risky, certainly for mass adoption and rely on electroencephalography (EEG) or reflected laser light to read neural activity. The problem with these non-invasive methods is that the data is limited and messy.

There is also the problem with the specific nature of these devices. It is not clear that they apply to more than specific uses or can even develop with the patient’s progress. There is also the speed of reaction problem. Such devices may be fine for deliberate action but to slow for fast reactions and decisions. Machine learning may help such devices adapt to patients and this work is already having some success.

Influence

Neuralink has set the pace in this field with proven progress. There are other invasive techniques, such as stents in major blood vessels and others such as Kernel are using non-invasive techniques to send light into the brain and read what comes back. This is a fast moving area that holds promise as human-machine brain interfaces develop. The possibilities have been explored in Clark’s AI for Learning (2020).