Until very recently, I thought AI was artificial insemination. Then I had experience of the unreal world of artificial intelligence and decided that it is corrupt, disingenuous and utterly untrustworthy. AI creates lies and I do not trust liars. That’s the short answer.
Here’s the long answer.
I was contacted by a PhD candidate, who asked me where she could obtain a book I had written. As I had not, in fact, written such a book, I told her so. “But you must have written it” she said. “It’s on the internet.” I was puzzled and ignorant. She told me she had found the information on ChatGPT. I investigated and discovered not only that I had written this book, but two other books, which I also knew did not exist, were listed about related subjects.
Then a university professor told me he had received an essay from a student who had included in her bibliography 15 (yes, 15) works that did not exist but which she claimed to have consulted.
READ MORE
So at one level AI can create “facts” that do not actually exist. At another, and perhaps more worrying level, someone makes use of those “facts”, thus in some obscene way validating what is, at the very least, a lie, a deceit.
I next discovered that a retired professor of accountancy, Richard Murphy, had a similar experience: an essay allegedly by him that he hadn’t written. Murphy had engaged in correspondence with ChatGPT, which admitted that the alleged work did not exist and that a “mistake” had occurred in what ChatGPT called “generating plausible-sounding article titles and publication dates”. There was no apology, just an acknowledgment that more care must be taken in future in this creation of “plausible-sounding” fakes – sorry, “mistakes”.
What does this tell me about AI-generated information? It tells me that ChatGPT, chatbot (which I had thought was simply a wine label – Chateau Bottled) and similar sources of disinformation are liars and should not be trusted even if they say that Christmas Day falls on December 25th each year. They may not have the intention to deceive, but the result is the same: disinformation, untruths, falsities.
At a time when we are continually being introduced to “the new”, it is very important to realise that artificial intelligence isn’t “new” at all, that we have had some form of computerised human wisdom available since at least the second century BC when the so-called Antikythera mechanism was built, described today as the world’s first analogue computer.
Fast forward to the 1950s when a computer known as the Parametric Artificial Talker (PAT) was developed at Edinburgh University with the capacity for human speech.
Although there were jokes about whether PAT could write novels, a serious warning had already been voiced by Alan Turing – largely credited with busting the German Enigma code in the second World War. In his 1948 paper Intelligent Machinery, Turing had questioned the entire ethical implications of “intelligent behaviour”; he even developed a “Turing test” to discover whether a computer could think like a human and replicate human behaviour.
If he had lived beyond age 41 – he died by suicide in 1954 – Turing would probably have argued against the development of what we now readily accept as books written by AI, on the grounds that they were ethically wrong, or, in his words, promethean. Today, the Alan Turing Institute in London does precisely that.
In the late 1960s the poet Lawrence Durrell, who had encountered PAT at Edinburgh in 1962, foresaw the coming of a “new age when the first book of poems by a defective and visionary computer” would be “launched upon an astonished world”. So it’s not new at all.
The insidious presence of AI is causing havoc in the publishing industry, which has become suspicious of anything that looks as if it might have been written by a computer. A book withdrawn because its purported author did not actually write it is not mere plagiarism. This is an assault on – dare I say it – our intelligence.
The idea that a company called Anthropic, which sounds like a benevolent relative, is scanning books in order to replicate them with bogus versions is repulsive to anyone interested in the truth. This is the third level of misinformation: when we entrust our credibility to a suspect entity we are opening ourselves to the likelihood of being deceived. Lecteurs complaisants?
There are quite possibly AI-generated books that are entertaining. I think a book written by a robot would be hugely preferable to some written by younger writers whose work I regard as lacking in literacy or grammatical cohesion. But I am old-fashioned.
If politicians are caught telling lies then we cease to trust them. If our partner cheats on us then we stop trusting him or her, don’t we. To accept AI and its lies into our lives means adding yet another layer of what is, in effect, self-deceit. It is, really, another form of artificial insemination.
Richard Pine is an author living in Corfu, Greece, and contributes a Greece Letter to The Irish Times
