AI, Education & ChatGPT

Dec 11, 2022

Hello from London!

This is the final regular submission of the year. It's been a pleasure sending you my weekly recommendations on reading, musings on life and what's next. Expect a couple of ad hoc emails over the coming weeks to the end of the year. And perhaps a few changes? A few people have suggested that I rename this newsletter "Rahim Recommends" or similar. But after nearly 6 years, change seems both inevitable and difficult. One thing is for sure: 2023 will be different. In the meantime, I hope you have a wonderful holiday period.

Like millions of nerds around the world, I played with AI this week. Both ChatGPT and LensAI. The latter interpreted me either as better looking version of myself or some sort of weird futuristic being. It was both fun and scary.

But today, I'd like to muse a little about ChatGPT and its application to education, the latest incarnation of AI for the masses.

Essentially, ChatGPT is an AI-powered chatbot allowing anyone with an account to have a simulated human conversation. It went live just over a week ago and while it is not quite the finished product, I'd say this is a game-changing moment in tech. It took just 5 days to reach 1M users. In context, it took Facebook 10 months to reach the same level. It grew so quickly that it's struggling to deal with the demand. It's one of those moments that help you envisage how the world might be reshaped.

ChatGPT uses language technology, an AI model from OpenAI that has been trained on a huge amount of information, e.g. the internet, Wikipedia, TV, books and other sources. It can answer simple, complex and follow-up instructions. Notably, for my friends in education, students, teachers and the like - it can write essays at the level of a university student. I'll grant you this: it's not perfect. It's also uninteresting. It feels like a conversational Google, but one that's not always correct. It's only as good as the data you put in, you see. That said I could quite easily see Google finally releasing its own conversational AI that had a heap of controversy around being sentient.

But what about that essay? How would it do in the context of a real exam. iNews got it to take a GCSE History exam and ChatGPT scored the equivalent of a low A grade (7):

Simon Beale, history and politics teacher at Vyners School in London, said the chatbot showed “more developed vocabulary than you would expect to see from the average GCSE student”, but that the answers were patchy, “hollow,” and at times showed a lack of understanding. ChatGPT also coughed up blatant errors, claiming at one point that Japan, Germany and the Soviet Union never joined the League of Nations. Mr Beale, who marked the answers, said he would have smelled a rat if a student had turned them in.

Or as one US humanities teacher put it: "What GPT can produce right now is better than the large majority of writing seen by your average teacher or professor."

I don't think ChatGPT will take the place of a tutor any time soon - but it may become a tool for students as they advance through their studies. The issue around AI models is that there is no incentive for them to have the correct answer. All it does is, initially, is provide consensus to what data it takes in or that it is trained by. It's a language model, not one that is based on fact.

Imagine a student then, who is mid homework, needing to check facts, info or data. There are multiple ways: check notes or the textbooks, use Google, or use a language AI model like ChatGPT which would give that consensus view. The choice might be quicker and good enough, if the student has some semblance in what they were studying.

And there are wider implications of how homework might change:

Imagine that a school acquires an AI software suite that students are expected to use for their answers about Hobbes or anything else; every answer that is generated is recorded so that teachers can instantly ascertain that students didn’t use a different system. Moreover, instead of futilely demanding that students write essays themselves, teachers insist on AI. Here’s the thing, though: the system will frequently give the wrong answers (and not just on accident — wrong answers will be often pushed out on purpose); the real skill in the homework assignment will be in verifying the answers the system churns out — learning how to be a verifier and an editor, instead of a regurgitation. What is compelling about this new skillset is that it isn’t simply a capability that will be increasingly important in an AI-dominated world: it’s a skillset that is incredibly valuable today. After all, it is not as if the Internet is, as long as the content is generated by humans and not AI, “right”; indeed, one analogy for ChatGPT’s output is that sort of poster we are all familiar with who asserts things authoritatively regardless of whether or not they are true. Verifying and editing is an essential skillset right now for every individual. It’s also the only systematic response to Internet misinformation that is compatible with a free society

The application to education is wide. Could this change the way we learn? Sure. If we think wider, could AI take your job. After seeing a glimpse of this, I'm sure that it eventually will.

These New Chat Bots Could Change the World. Can You Trust Them?

Stay Curious,

Onward! - Rahim