With this blog, we are kicking off a short series of articles on AI in education to explore what the concepts of intelligence and learning mean in the context of AI.
When thinking of artificial intelligence (AI), it is often too easy to think of dystopian films where intelligent robots have taken over the human race.
But authors and filmmakers are not researchers, nor are they EdTech entrepreneurs. Where the former will get ahead of their imagination, the latter look at the reality of AI, its potential but also its limits. Is AI actually intelligent, and if so, in what way?
The term itself was first coined in 1956 by the computer scientist John McCarthy, who defined AI as “machines that have the ability to achieve goals like humans do”. He predicted they would do this by thinking and acting “humanly” and rationally.
The first AI systems followed a set of pre-programmed rules, but it soon became apparent that this wasn’t sustainable because humans don’t “work” in that way – they learn, adapt, socialise, feel and respond, sometimes irrationally.
Early innovations developed between the 1960s and 1980s saw the emergence of three conversational programmes using imitation of real human conditions. But these also followed pre-programmed logic and were unable to respond to complex human statements with anything other than fixed phrases, such as “very interesting” or “please go on”.
More recent AI is based on machine learning and finding statistical patterns in large banks of historical data. These modern machines will only replicate human intelligence by accessing high-quality data relating to a problem, as well as the algorithms to empower it to learn from this. To do so, it must also contain enough processing power and memory.
Having this single, non-transferable skill doesn’t make it intelligent in the human sense, however. Even then its “intelligence” is limited. It wouldn’t be able to make decisions on instinct, emotion or assumption the way a human can, for example. But, because its use of statistics makes it consistent, it may beat you at chess – or a game of Go, as the (human) world champion found.
Many of the companies we work with on the UCL EDUCATE programme build AI into their products as they seek to offer personalisation of learning. It is vital that they understand the potential, and limitations, of AI to be better able to make the right decisions for their product and ultimately, for children’s learning and teachers’ teaching.
For further information, download the first of our Byte-sized edtech research document on AI.
You can read the two other blogs on this series here: