by MERLIN CROSSLEY

In many ways ChatGPT looks like a “Siri for everything”. A universal reference source that will help learning and eclipse previous repositories – the Oxford English Dictionary, the Encyclopedia Britannica, Wikipedia.

But I keep thinking ChatGPT is different. Not just that it might be like Space Odyssey’s HAL 9000 and ultimately prove dangerous and threaten humanity, but that it may drive things towards homogeneity, and we’ll end up with a world that isn’t just average, but very average. One that reflects what people call “the lowest common denominator.”

Students are already using ChatGPT as a tutor. I use it too, both to check details I can’t quite remember, and to explore topics that are new to me – like a student might. When it comes to new things, I am never quite sure whether to trust it or not.

To me the important question is – how will it manage to elevate quality above quantity?

Doing this is vital to scientific and cultural growth.

It pays to ask ourselves how any of us currently balance quality and quantity.

I think the process works like this.

First experts, peer reviewers, critics pass judgement, and together they convert quality – which is an elusive and sometimes fraught concept – into quantities that everyone can agree on, be it hats for restaurants, stars for films, a Booker prize for a book, or an Archibald prize for a portrait etc. Pioneering experts launch quality works on their way powered by a few indisputable quantity points.

Then after the experts have dispersed, time steps in. Quality items stand the test of time. If something has gravity then quantity proxies keep sticking to it, things like citations for scientific papers, and contentious but sometimes useful, book sales and box office sales for films and theatre, inclusions in gallery collections for art works.

When we design a syllabus and teach, we, as the experts, also take great care to put quality works in front of our students. Some universities prioritise teaching “great books.” There is no strict definition of what a great book is, but it is generally one that has won over the critics and passed the test of time. A great book is one that is worth preserving.

And the first books were hugely expensive, so they were carefully selected. Scribes copied out classics and parchment texts were stored in libraries, monasteries, and the homes of the wealthy. After the invention of printing, books and knowledge in general became more readily available, but there was still a cost. And books take a long time to read.

So these are the three hurdles that separate works of low and high quality: winning over initial critics, standing the test of time, and being worth the sustained investment.

But all that changed with the internet. The experts were outnumbered, the time delays were gone, and the costs were minimal.

Newsfeeds now vary in quality. Even major newspapers often publish questionable articles. Consumers of news are being split into self-reinforcing bubbles or echo chambers. Many people believe this poses a risk to the very coherence of an inclusive society.

Now what about ChatGPT and its friends?

These language models don’t know anything, but they can make predictions based on what is common, having fed on the text on the web. Everything depends on what dominates the web. If the quantity of information grows exponentially then new things will quickly overwhelm older things. This is quite different from university libraries where old books are treasured.

Interestingly, the first chatbots have apparently been trained not on the internet as a whole, but on collections of texts that have been collated, including Wikipedia but also reputable websites such as The Conversation. The idea of continuing to curate data sources is appealing but who will judge? My fear is that over time there will be no selection and chatbots will feed indiscriminately.

It could be that ChatGPT is currently as reliable as it will ever be – and even now we know it hallucinates!

If academic teachers now have a new role – to equip students to manage in a post-ChatGPT world and ensure that they can use these tools wisely and effectively – then we have a lot of work to do. Do academics have a bigger role than ever as guardians of quality? And if we start asking ChatGPT to help with our teaching, then who guards the guardians?

An interesting ride is just beginning. Perhaps a longing for quality will drive the emergence of trusted authorities and make them more important than ever, or perhaps pulp fiction will take over. Up till now there has been a balance, perhaps that will continue, but we should all be aware of the risks and do our best to respond to prevent a slide into intellectual mediocrity.

Professor Merlin Crossley is Deputy Vice-Chancellor Academic at UNSW SYDNEY