06/09/2025
Academia won't grasp AI until it recognises that it's not one skill. Mastering it requires a mastery of multiple skills all at the same time -- which is impossible challenging when many kids don't have a solid grasp of the fundamentals now.
Building smart chatbots, curating datasets and implementing clever input prompts is only a fraction of the educational problem that needs solving.
Mastering AI (is in my opinion only) - a stack of skills - that have never really been taught as a single set before.
The skills are as follows:
1. A solid foundational grasp of what different AI models can do, how they respond (reliably or not) to various inputs. Without this base understanding you wont make much progress and very few academics (or people) are deeply across the capabilities we now have
2. The skill to write software—whether directly yourself, through a prompt, via an AI agent, or even "vibe coding." Prompting alone won't give you the precise control needed to control the AI effectively. AI is only powerful when you master how to control its actions agentically. The return prompt alone has limited utility that is most commonly scrutinised.
3. The ability to define problems from first principles, no matter the topic. If you can't break down a problem step by step, you won't know how to use AI to advance whatever you are trying to achieve.
4. A hypothesis driven mindset that's ready to iterate problems, critically analyse results, and validate them. This is rooted in the scientific method and to some extent the entrepreneurial mindset of validation. Both divergent skills.
5. Accurate and pursuasive communication skills and understanding two ways: (1) Describing exactly what you want to the machine to do to maximise the accuracy and precision of results, and (2) having an ability to communicate results to humans in clear, persuasive, and actionable ways. Both hopelessly divergent and complex skills.
And
6. You still need adequate DOMAIN knowledge in whatever topic you're working on. If you don't understand the underlying topic and concepts, AI won't help much.
That being said - with proper training, you can use AI to quickly fill knowledge gaps, and speed up your learning process. Mastering skills 1-5 will be a massive advantage to those trying to master the domains of 6.
But then there's the X factor.
The Mindset of the Student.
Do they have the agency and confidence to master the skill stack - and will traditional teaching processes get them there?
I'm not confident.
Why?
Zero Teachers and Academics were taught that stack of skills. It's like a Unicorn skillset.
I think Teachers will become (eventually) like coaches. But in a way, the teachers need to go back to school too.
Add into that the Bureaucracy burden of change -- I don't see it happening systemically for some time.
If anything you probably need a scenario where the teachers are collectively educated by a superset of new technologists in the new world.
Not sure how that'll go down though.
Nope, I think the best thing to do is teach young people. They'll figure it out from first principles.
The super students are coming.