Gen AI teaches us something important about learning and it’s nothing like you think.

It’s been some year building upon the previous work of AI-storytelling. ‘Chairman’ showed how to recreate archival films using academically approved approaches and was shown at various summits.

Empowered commissioned by Dr Carlton Brown depicted Black women breaking glass ceilings. It would show on cinema screens in the US at the first AI film festival. ‘Ai-casting’ developed visualisation for my authored BBC international radio reports examining how 2025 will likely see a shift in podcasting to ai-casting.

News this week of the Martin Scorsese Virtual Production Center at NYU launching an AI film course will galvanise AI in education storytelling and journalism for 2025. But that’s not the story here in this post.

The expression ‘data is everything’ is common place in AI, but it masks a broader learning model largely not implemented in secondary and tertiary learning systems.

Notwithstanding the ethical concerns of copyright, what’s made Generative AI models, such as Chat GPT extraordinary in their use compared to AI stems from various algorithmic architectures. These include transformers (providing nuanced interpretations), decoders, computing power and data.

Modern theories of learning generally in the West are based on procedural learning. This is similar to AI’s widely used supervised learning which up until recently was dependent upon labelling data it was ingesting. Before 2017, a pivotal year for Gen AI (discovery of transformers), AI systems generally yielded results via data people had codified.

By way of an analogy, say for instance you had a group of students learning journalism you’d essentially feed them with labelled terms e.g. objectivity, package, inverted pyramid etc and books with journalism.

Then let’s say you had a second group. With them you dropped them into an environment for learning storytelling. It would include cinema, literature, advertising, ways of seeing, freudian concepts, with Business Ai etc — a fusion not labelled or directed for use. The students here in solving their storytelling conundrum would employ ‘making connections’.

They’d be looking what’s relevant and isn’t from discerning patterns.

This is how LLMs like Chat GPT work. Their vast computing power enables them to decipher huge amounts of data in seconds and produce results.

What if secondary and tertiary education adopted this model of learning which has proved for AI to be transformative and impactful? Could Gen AI’s model be one for learning institutions?

Colleagues and I piloted two programmes over the last eight years that did just that. Cinema journalism and the LAB both involve collaborative ‘making connections’ learning adopting Gen AI’s framework. You can read more here.