Artificial Intelligence is not as smart as it ‘thinks’ it is

Creat claims, as by now we are all too well aware, are being made for the future of AI – the internet development of artificial intelligence which it is said is about to transform, not just the world, but the cosmos.

Though there are Jeremiahs who are filling the airwaves with their doubts about both the system itself and the huge financial and physical resources being heaped on it, as yet all without any profitable return. We are urged to have faith, for all will be well in the best of all possible worlds.

Yes, we have heard this all before, with some recalling not only the great Dot Com bubble in 2000, but also the startling events of the autumn of 1929, when the financial system of the USA simply imploded taking with it the economies of the USA as well as plunging the wider world into economic difficulties. Economic historians are aware of similar giant failures in the past, going back indeed to the ancestors of them all, the Tulip Mania and the South Sea Bubble. Some people learn from the past, but most people lured by the promise of easy money or money for nothing as with crypto-currency, don’t. Yet again the AI world is simply ‘too big to fail’ we are told, so it won’t fail.

I am not on the whole a person absorbed, as so many of one’s friends and relations are, by the internet. I certainly trawl the news morning and evening from Ireland, the UK, USA, Canada and France – which is more than enough to spoil one’s sleep at night.

But for the rest of the time, I am devoted to my own various literary projects (along with, of course, the composition of these pages). But one cannot avoid AI: it gets in everywhere. Ask for anything on the internet and the first thing that pops up is an AI response.

Yet I have found that one must be wary of these quick answers. In many circumstances I have found the AI response was simply wrong or inadequate. Certainly, it was not to be trusted.

In connection with one of my projects which involves Argentina in the early 1920s, I recently picked up a second-hand university level book on the physical and economic geography of South America.

The author’s name, D. C. Money, was unfamiliar, so I queried it, adding the term geographer.

What came first from AI was a reference to one to the District of Columbia, where Washington is, and to money itself. This was wrong and useless.

I discovered the author’s full name, David Charles Money, and searched again emphasising the term geographer. Again AI assured me there was no such person known as a noted geographer.

In fact, only by resorting to the more humane Wikipedia did the answers his full name and career become clear. He was head of the geography department at an English public school and the author of many books.

Again, there was another instance. I wished to discover something about the former home of the Spanish novelist Vicente Blasco Ibanez in Menton in the South of France. I put in the search term Ibanez (which rang up a brand of guitars). “Ibanez Museum Manton” brought the AI information that there was no such museum. Instead, it offered a short list of other museums in the region.

In fact, the house in which the novelist – a political refugee from Spain at the time -lived during 1922-23 is now open to the public with a display of materials relating to Ibanez and his work, which is used also as a cultural venue. There is also a much admired garden he created. He was the author of Blood and Sand and The Four Horsemen of the Apocalypse, greatly famed in their day and still widely read in Spanish speaking countries.

AI is based on a dredging of masses of internet data, so it can claim to “know” a great deal. But it is useless for students at school or university, as it provides no specified sources for school projects or exact reference for college papers, let alone higher-level theses. Legal references it provides have proved in the past to be inventions.

I could go on – and relate other instances of the AI’s over-literal approach to queries, its total lack of flexibility, and American cultural overload. I will stick with Wikipedia and wait for the AI bubble to burst.