Many of us remember the classic old tale of Cyrano de Bergerac, the French play written in the 19th century, which was based loosely on the 17th-century French novelist and playwright. The play has been rendered into film several times; if you're a film buff, you've likely seen one or more.
In the play, Cyrano is a daring swordsman and accomplished poet, saddled with a coarse appearance capped by a prodigious nose. Christian de Neuvillette is a dashing, handsome, and yet inarticulate young soldier, who recruits Cyrano's help in writing notes to win the heart of the winsome Roxane. Cyrano tries, writing letters signed with Christian de Neuvillette's name, but the letters are a deeper fraud, as Cyrano is falling in love with Roxane himself.
Now, the rise of artificial intelligence (AI) may be giving rise to a new generation of Cyrano de Bergeracs. Over the U.K. Daily Sceptic, author Joanna Gray has some interesting insights.
At a recent training day I was taught how to use AI speech recognition to write reports, letters and emails. A teaching assistant led the session:
English has never been my strong point and I constantly make all sorts of spelling and grammatical errors. AI has changed my life. If I have to record an incident about a pupil I now use the microphone button on Gemini, explain the incident, ask it to rewrite as a decent school incident report and wait a couple of seconds. I then cut and paste Gemini’s response and send it off.
I know and like this particular woman but she is not one of life’s natural communicators. She prefers saying “I done that” to ‘I did that’, for instance. She also has a wicked sense of humour. I read the report that she done – it was improbably immaculate.
Many of us who are in the wordsmithing business have our own particular styles, and experienced editors are familiar with the distinctive voices of the writers they work with; it is not at all unusual for an editor or reader to be able to recognize a particular author they are familiar with, without bothering to look at the byline. But AI has the potential to muddy that water.
Here's the complication, according to Gray, and note that she is channeling Cyrano de Bergerac as well:
What makes AI voice technology a different issue from that of mere plagiarism is the fact that the kernel of the messaging does indeed emanate from the person commissioning the AI. It’s all about the prompts. Christian did love Roxane, the teaching assistant did experience an incident with a student that needed recording. Friends and family all use the technology with pleasure. I finished off my online training by asking Claude the answers to everything and ticking the relevant boxes in another window.
It's an interesting problem. It's not plagiarism, but neither is it a true assessment of the knowledge and communication skills of a writer. Now, some authoring tools can take a finished document and examine it for spelling and grammar errors; this is nothing new. Word-processing packages like Microsoft Word have had spelling and grammar-check features since the very beginning, and those could be loosely described as an AI tool. The newer, more sophisticated tools, used this way, remain simply tools.
The problem comes when AI is being used to replace original work, and it's a growing problem in academia and business. One report indicates that 92 percent of university students worldwide are using AI to generate assigned work, while 95 percent of the academic community is concerned about the use of AI to produce work on which students' grades are based. Needless to say, that's troubling, at the very least.
On Monday, I wrote about how AI was replacing Kenyan cheating-industry workers as a way for students to commit fraud in their work. Now it seems the problem is larger than we may have thought.
So, how to counter these modern-day Cyranos? To employ a phrase that Cyrano himself may have understood, instead of setting the fox to guard the henhouse, perhaps the answer is to set the fox to catching other foxes. Use AI to detect AI.
After years of disappointing results, multiple firms now claim that software can reliably distinguish between AI-written and human-written text. One is Pangram Labs, the New York City-based start-up that makes Pangram. “Detect AI-generated content with 99.98% accuracy,” the firm says on its website.
Scientists and research organizations are among those using the tool to spot AI’s traces. One in eight biomedical articles last year contained some AI-generated text according to Pangram, a study reported in January. In June, the premier computer-science conference NeurIPS announced that it rejected 18% of submissions after screening them with Pangram. Users of the preprint server arXiv can now check Pangram’s verdict on any article there, by visiting a mirror site called alphaXiv that has installed the tool. And the University of Chicago in Illinois says that it has started using it to vet students’ coursework.
This is key. AI, the tool, is not going away. Its use will only increase in academia, in publishing, in business. And the detection software will only improve. My wife, who runs a small publishing company, has experimented with several tools, and she now routinely runs submissions through one or more AI detectors to examine them for generated content, and she has rejected manuscripts found to be AI-generated. This is how we must deal with this moving forward, because genies are, historically, very reluctant to go back into their bottles.
Oh, and as for Cyrano de Bergerac's movie adaptations, if you haven't seen the 1987 Steve Martin film Roxanne, I can't recommend it enough.
