The chatbot is amazing and a menace (and I DID write this myself!)

WHEN I sit down to write this column, I try to explore a topic that I care about or really interests me. Perhaps something that has come up in conversation at the dining table during the week, or is a subject of discussion among my friends.

Sometimes it’s easy and I have a burning issue I want to get off my chest. Other times I’ve had a boring week, or I’ve been too busy to keep an eye on current affairs, or the kids were sick and life involved wiping snot - and I spare you, dear reader, from a rant about the price of Calpol or how many Disney movies a small human can watch in one day.

In those uninspired weeks, what if there was an easier way to knock out 900 words of reading gold?

Last week, a director I’m working with sent me a one-page proposal about a documentary telling the story of the laying of the transatlantic cable from Valentia Island in Kerry to Heart’s Content in Newfoundland. It was a great programme pitch, promising intrigue, invention and mad entrepreneurialism. As I read it, my stomach dropped because myself and the director are currently halfway through filming a documentary about the very same topic.

You can look forward to seeing our offering at the end of the year, but who were our rival storytellers? Turns out the director had asked the artificial intelligence language model Chat GPT to “write a proposal for a television documentary about the laying of the first transatlantic cable”. And the AI model hit it out of the park.

The proposal was full of historical detail, from the human characters involved in the endeavour to the engineering and scientific obstacles overcome.

Relieved that our forthcoming documentary wasn’t about to be eclipsed by another, I was intrigued about what else Chat GPT could do. If this AI model can write an extremely convincing documentary proposal, could it write a 900 word column on a Tuesday for The Echo?

Up to now, you might have encountered old fashioned Q&A bots on websites that are supposed to answer your queries but are mainly there to stop you dealing with a real life human. What’s special about Chat GPT is you can ask for the answer in a certain style: a speech, an essay, a newspaper article, or something Shakespeare worthy.

I asked ChatGPT to write an essay in the style of Shakespeare about the Shandon Bells.

“Hark, fair reader, lend thine ear and hearken to the tale of the Shandon Bells, a chime of bells that doth resound through the city of Cork with a melodious tone that hath delighted many a heart and soul.”

The model is really impressive when asked to describe something factual and well established like a scientific principle or historical event, but it has limitations because the model learns by reading and processing trillions of examples of text and then it regurgitates and rehashes those samples.

If the model hasn’t ‘learned’ anything on your chosen topic, it just makes stuff up! It can be very convincingly presented, but the reader needs to know if the information is correct or if the ‘essay’ produced in 90 seconds is nonsense!

I asked “Write a newspaper column about the Cork hurling team for the daily newspaper The Echo and this was part of its offering:

“With a blend of experienced veterans and talented young players, the team is a force to be reckoned with on the national stage. Led by manager Kieran Kingston, the team has been making steady progress in recent years and looks poised to challenge for major honours in the coming seasons.”

Not bad, except Pat Ryan is the manager!

Some people are hailing ChatGPT as an invention as important as electricity in transforming the way that we will live in the future. From writing computer code to offering personalised medical advice, to translating languages, it is “the next big thing”. Others are worried that it will compromise human creativity, usurp jobs and become an unregulated technology with untold unforeseen consequences.

Apart from the accuracy issues and the ‘computers taking our jobs’ concerns, there are other worries. The energy required to build and support this AI ecosystem is enormous. Every time someone goes to chat.openai.com, those searches and replies are built on staggering amounts of computing power and IT infrastructure. Actual physical computers, enough to fill football fields, requiring lots of power to run them and cool them. When we talk about building more data centres, it is to enable technologies like this to expand.

If these large language models are going to become core technology, it is going to suck up a lot of the world’s resources.

In the midst of an energy and climate crisis, when we dramatically need to curb our usage and move to renewable energies, you have to wonder if this technology should be a priority for the human race at the moment.

The other major problem is that it doesn’t really matter what I or you say, or even what governments and states say, this technology is being driven by big, powerful, transnational tech companies. We should pause to think about the power this technology promises to wield, being concentrated in a handful of global corporations.

Given the potential enormity and wide-reaching ramifications of its various applications it almost seems too big to regulate, the industry is proposing self-regulation to keep the tech in check.

In Ireland, we know all about how self-regulation of big corporate entities can translate to little or no regulation and how citizens suffer the social consequences.

So, was this column written by me or ChatGPT? Go to the website and ask the model to ‘Write a column for The Echo newspaper in Cork about ChatGPT’ and compare...