The Educational System is a Startup gone wrong.
What's the most optimal way to use my time in the next 5 to 10 years?
Definitely not going to university, not because university doesn't teach you anything, but it's more about the fact that I know I can achieve much more by myself. I also think that transitioning to a world post-AGI, we have to focus more on developing agency rather than extanding our knowledge. I'm not saying knowledge won't be necessary, but it will become a commodity. What does achieving more by myself mean? Can I truly learn more outside of University?
I think in order to learn optimally you need to:
- Know what to learn
- Understand the KPIs
- Have effective feedback cycles as often as possible
In some way, the educational system manages to get none of these right... I mean, someone thought about a model of "what teaching kids should look like", then tried to make our human brain fit that model. In startups, you don't start by saying "here's a cool idea, let's try to push it down people's throats", it doesn't work most of the time; this is usually called a solution in search for a problem. You have to find a problem and build a solution from there. And maybe I'm wrong to begin with because the core problem the educational system tried to solve was not to educate but to make obedient citizens and cherry pick those who would prove most useful to society. My point is that if we want to come up with a better solution to educate people, we must start by understanding how does the human brain work, and make a model that leverages our inner workings. I'm not making any claims that I have any insights into how the brain works... But I don't need to, because other people smarter than me came up with transformer models, which mimick the way our brain works.
What are some key characteristics of training a Transformer?
- Curated dataset
- Clear feedback, that measures how close we were to achiving our goal
- Iteration upon iteration
What should we adopt from the way AI learns and use it ourselves? I think we focus too much on this approach of providing the students with all this true and valid information, then making them memorise it and apply it. We modelled the system as if brains were relational databases... They're not! We could learn like AI, by trying to perform a task, measuring how well we succeeded, changing our internal parameters and trying again. This is way more efficient because you create additional relations/connections between the data you store within your head, based on how it's applied, in which context, why something doesn't work, etc. You also learn a lot and develop intuition once you understand why something cannot be the way you want it to [for example understanding why a coil does NOT have impendance in a direct circuit *(except for when you turn the switch on or off) makes you understand more about AC and DC and coils and electromagnetic fields at the same time] I could probably go into why this method of learning is more efficient and make a lot more arguments for it, but I'll leave that for another essay, now let's just accept it as it is.
How would this translate into the real world?
- Do first, not tell (let them figure it out)
- Fail → feedback (learn from mistakes)
- Choose your goal (intrinsic motivation)
- Project synthesis (see real use cases)
- Personal "Why" (set your KPIs)
So, I think education doesn't have to be boring, kids shouldn't hate going to school... In one of my next essays I'll challenge some of the beliefs we've had about the educational system for decades.