When I was bright young computer spark, in the eighties, in the early days of “Not so super intelligence” we were building expert systems, which were confidently predicted to take over from your family doctor, your bank manager and other professional advisors, the human sitting at a computer and entering your answers to its questions. This, as we know, never went anywhere, and the stupid cry of “Computer says No!” rang through the land to much derision.
I fell foul of this myself in the mid nineties when, newly arrived in England, I tried to open a bank account. Computer kept saying No! until I triumphantly produced the cheque book of my previous English bank account and waved it at the bank manager, who quietly overrode the computer and opened us an account.
There have been a series of moral panics over my lifetime – the expert system future that didn’t was accompanied by the fears over television use (“Amusing ourselves to Death” 1985) and the dire effects of computers on human health and intelligence. Before that the typists and the double-entry clerks fought a battle to save their jobs. We even had a system in the early days of e-mail (1978) when the secretaries solemnly printed out emails and delivered them to the bosses’ desks.
I recognise that the tools and systems we were experimenting with back in the eighties were child’s play compared to today’s scary monsters. And the moral and ethical dilemmas posed by the new tools are much harder to negotiate, but they are not new. Plagiarism and misinformation have always been with us, but we have gotten quite good at spotting them.
Not So Super Intelligence
Outright lying in print or speech is as old as literature and rhetoric. Fact-checking crept into historical writing at snail’s pace –the Venerable Bede, when recounting the miracle-packed martyrdom of St Alban, tried to check his facts – but only changed the name of the river in the story from Thames to Ver.
As a poet, I have used Roget’s Thesaurus, Chambers’ Big Red Book and a site called Rhymezone all my life to provoke my poetic imagination. I almost never use any of the words or rhymes suggested, but these word-hoards act as a stimulus. How, I wonder, is this different from asking an AI agent to write a verse or two, or a purple passage?
The central moral and ethical issues with using AI to do work for you boils down to two things: am I lying when I put this out as my own work? and: Who am I depriving of a legitimate livelihood?
These are completely different kinds of question, but to some extent they converge, as we will see.
Some cases are open and shut. If a preacher stands up and delivers a sermon that has been partially or mostly written for him by an agent, most people would regard that as a moral failure. A sermon, in our era, is defined as something that comes from the preacher, as his or her belief. If he generates it – and some do, I’m afraid – it is effectively a lie.
Is this different from a preacher in times past reading from a book of homilies? Yes, I think so. The existence of such books of homilies were well known, and the clergymen did not attempt to pass the sermon off as his own. He was, after all, reading it from a book (unless he were a complete fraudster, and copied it out in his own hand to make it seem his.)
Not So Super Intelligence
Sermons are at one extreme – but what about a lecture or a general speech? Is it morally wrong to deliver a lecture or a speech that you haven’t actually written yourself? Not morally, though it might land the speaker up in a mess if later asked questions or put on the spot. Ethically, though, it’s a bit different. The speaker who knows their subject so poorly that they have to generate their speech is something of a fraud.
The speaker might put up the defence that they know the subject, and the opinions are all theirs, but they don’t have the skill to write it up into a coherent lecture. This sounds convincing, but as every writer will tell you, it is the way you put your thoughts into words that is the key to communication. And public speech is communication, pure and simple.
There is another reason why one should write one’s own material, rather than generate it. When we write, we think. We are exercising the brain, which is even more important than exercising the body. It we have to pause and think of the mot juste rather than taking the next phrase suggested by some engine, we are keeping the vocabulary engine fit, and also, we will produce something that sounds like us, and not a characterless bot.
Not So Super Intelligence
The obvious comparison are the generated images we see everywhere on posters and cartoons and posts. We – so far anyway – can spot them right away, and they create a tiresome sameness. It only takes a momentary glance at a cartoon by Ronald Searle or Leung or Gary Larson to spot the difference. Doubtless bots will be mimicking those greats before long, but if Claude’s attempt to write a poem in the style of Imogen de la Bere is anything to go by, it’s going to be a steep climb.
The other moral question we have to ask when using AI to produce things is this: who are we doing out of a livelihood? Illustrators already have a tough life, as do editors and technical writers. By doing them out of a job, are you not condemning us to a world of sameness and dullness, and them to worklessness, after perhaps a lifetime dedicated to producing the excellent work, on which the bots have been trained?
There is also evidence that the bots get stupider over time, as they feed on their own – often inaccurate and incomplete – information. I have been asking an AI bot the same question for some weeks – who wrote the first two lines of the doggerel it produced in my name? The answer each time is different, and gets less and less accurate and useful. This tendency to feed on its own inanity and misinformation may be corrected in time, but the requirement on us to weigh our usage in the moral and ethical balance will never go away.