Tuesday, June 22, 2021

Monkey’s mind plays pong - no hands. Frictionless interface. Here’s my musings on 10 possible, and probable, mind-blowing applications in LEARNING

This is the most fascinating video I’ve seen in 2021. It’s haunting me. Not because of what them monkey is doing, that’s astounding in itself, but what it points to in the future, that’s mind-busting.


What’s happening here? 

The monkey got  banana milk-shake rewards through a tube by learning to play ‘pong’ with a joystick. He got rather good. Remember that pong is a complex set of skills, where you have to anticipate the physics and movement of a bouncing ball off fixed and mobile objects. It requires good cognitive and motor skills. Then they pulled the joystick lead out of its socket and  simply read the data from a fine fibre array in the monkey’s brain using AI to translate those thoughts into action. They are literally using AI to mind-read. If this doesn’t blow your mind, what will?


So what are the potential consequences for learning?

1. Learn to move

Neurolink, who developed this technology, the robot that inserts the fibres, the fibre array and the AI that reads the brain signals see its immediate application in helping those with physical disabilities, missing, damaged or paralysed limbs. One can literally use your mind to operate robotic arms or legs. So we’re on the move.  But let’s also speculate on including those with dyslexia, dyspraxia, autism, ADHD. Could we also tackle issues of inattention, demotivation and persistence?


2. Learn by doing

Beyond the fantastically inclusive ideas that this enables disabled people to perform physical tasks in life, it also means they can ‘learn’ fine-tuned motor skills. You could play a musical instrument, conduct an orchestra, paint, play lots of sports, computer games and build things. You can learn to DO things. Note that learning to DO things can include doing things at any distance across the internet, even for able-bodied people, just by thinking them.

This is the instantiation of a famous experiment that took place 20 years ago by Pascuale-Leone (2001), where two groups practised piano, one on a real piano, the other just isn their head. Both groups resulted in similar brain changes as measured by FMRI.

Let’s take this a step further, we could ALL practice DOING things with AI software that formatively assess us, in realtime, provides coached feedback to improve our performance. This feedback can, of course be automated. Such coaching would be personalised, available at any time to anyone at any place, anytime. Behavioural skills could also be assessed. Want to know if I can do something. I need only think it and you need only interpret and assess it.


4. Thinking skills 

This may also apply to other cognitive skills such as speaking and writing. We are moving towards a world where we do not have to use a physical interface, keyboard, mouse or touch screen towards voice, already available in our cars and home on our phones, with voice activation and voice assistants, such as Google Assistant, Siri and Alexa. This will move towards invisible interfaces straight from brain to screen or machine. We all have a phonological loop. Close your eyes and say ‘I am Scottish’ in a Scottish accent. You can do this because you have an internal voice as a feature of working memory. That can be read and with AI translated into words on a screen or speech.


5. Tricky skills

Maths is an area of catastrophic failure. The vast majority never get beyond basic numeracy, even then, calculating an average or understanding a logarithmic scale on a graph is beyond many. You can’t progress in maths unless you master pre-requisite skills, as it has high dependencies on prior knowledge. Imagine a system that reads what you think, diagnoses the misconception, corrects that misconception through feedback, worked examples and so on, then asks you to try things within or just beyond your level of competence at that exact moment. Personalised, adaptive maths, globally scalable from the cloud, could massively accelerate and increase attainment and massively reduce early drop-out.

Let’s take another different skill, learning a second language, another area of catastrophic failure. All kids in the UK and US get years of language education, yet few could ask for a sandwich in that language, never mind hold a basic conversation. Imagine being able to try that language, privately, without public embarrassment and have your thoughts read, interpreted by AI and constructive correction and feedback, say correct pronunciation, position in sentence or endings. Personalised, adaptive, language learning, with immersive practice available, to everyone at any time, would revolutionise language learning.


6. Critical skills

Now imagine cognitive skills, where you can engage your mind in conversation with an expert, who can tune there reprises to your level of skill. Now, transformers that encode and/or decode natural language, such as GPT-3, BERT, BART,  MegatronLM, Turing-NLG, We saw an interesting example some years back when IBMs Watson debated with a national debating champion. These transformers can generate text and start to show how complex interactions with humans, drawing upon huge databases of existing text. Digital Einstein is a recent example of this. Imagine being able to just conjure any famous writer, scientist, artist, whoever, and start asking them questions, in your mind. Your questions will be interpreted and answers based on their archived works, used to provide meaningful answers.


7. Translations

Learning is often hindered by the lack of available teachers and resources in m minority languages. Text to speech and speech to text are limited to mainstream languages. Facebook has recently managed to get AI to learn how to translate any language just by exposing the software to speech in that language. This would free all of the above and liberate those who feel trapped an unsupported in their own first language. 


8. Write to minds

Now let’s take a final leap. And it is a big leap. Almost the above is about reading minds, what of we can write to minds. This is already there with neural ink and other invasive technologies, but not at the semantic or skill level. Imagine being able to pick up maths or a second language very quickly in days or weeks not months and years. Wouldn’t that be desirable? Hell yeah.


9. Mental health

Let’s take another leap and suppose that breakthroughs allow us to write to brains in a way that reduces or eliminates depression and other widespread disorders. Think for a moment about the vast reduction in human suffering that would follow. There is no doubt that education has been increasing mental health problems, inducing stress and in some case suicide. Almost every major institution has killed or helped kill young people. That’s sobering.


10. Anything, anywhere, anytime

You may be thinking, could ALL learning be liberated by such technology? The answer is probably yes. Time will tell and we don’t yet know how long it will take. One thing we do know is that these innovations are now coming thick and fast. Combined with other physical, global innovations, such as Starlink, a global network of low level satellites that give high speed, 5G, internet coverage to the whole world with no blindspots, and we have an educational network that excludes no one.


Conclusion

They used to say that information wants to be free. That’s done. Most information. Is free. We should now ask whether education wants to be free. Teaching is a means to an end, the relatively permanent change in the long-term memory of the leaner. If we can automate that process, make it frictionless, personalised available to anyone, at anytime, anywhere, we should. Learning longs to be free. Let’s not let our conceits hold that back.


Bibliography

Pascual‐Leone, A., 2001. The brain that plays music and is changed by it. Annals of the New York Academy of Sciences930(1), pp.315-329.


No comments: