Skip to content
Home » How To Master The Art of Writing With AI: Alice Delorme Benites (Transcript)

How To Master The Art of Writing With AI: Alice Delorme Benites (Transcript)

Read here the full transcript of Professor Alice Delorme Benites’ talk titled “How To Master The Art of Writing With AI” at TEDxZHAW 2024 conference.

Listen to the audio version here:

TRANSCRIPT:

The Need for AI Guidance

You need me. You need me to save you from a horrible future. Now imagine for a moment that you work for a company that sells Swiss chocolate. In a couple of years from now, a very rich foreign country will open up their frontiers and decide to massively import Swiss products.

So this will be your time. You will run to your AI assistant, and you will ask it to come up with a good slogan and a nice picture for your marketing campaign. And here we go: “Indulge in elegance, Swiss precision in every chocolate moment.”

It sounds good. It looks delicious. So now you are ready to enter the market, along with all the other competitors who also use ChatGPT as their writing assistant. As a result, nobody in that foreign country is able to tell the different products apart.

So they end up buying the cheapest one, which is not yours, so you lose your job. Now you have to apply for a new job, which you do by going to your writing assistant, ChatGPT, and asking it to come up with a good application letter. But this time, you’re a bit wiser. So what you do is to decide that you will change the letter a little bit.

You will make it more personal, more individual. Luckily for you, the letter is quite convincing. It’s really good as is. So you end up just changing two adjectives and one verb.

And off we go, you send your letter to the recruiter, along with a hundred other applicants who also use ChatGPT to write their letter and who also change just two adjectives and one verb. Well, needless to say, you don’t get the job.

The Dystopian Future

I could go on. I could go on and tell you about the next national election, where all the candidates would campaign with the same slogan and want to make Switzerland great again.

But I think you get my point. My point being, this is not the future that we want. Really not. This is a nightmare. This is George Orwell’s 1984. So we don’t want that. Now how do we avoid this apocalypse of boredom? Should we stop using AI?

No, of course not. We didn’t stop using cars the minute we found out that we could have accidents. What we did is to come up with rules, regulations, a code of conduct. We have people taking driving classes, taking driving exams to make sure that everybody who is behind the wheel is actually in control.

ALSO READ:  Jaron Lanier: How We Need to Remake the Internet (Full Transcript)

And this is what you need. You need a driving instructor. So welcome to your first AI communication driving class. The class will be composed of three lessons.

Lesson One: The Priming Effect

Lesson number one: the priming effect. The priming effect is what happened to our Swiss chocolate vendors when they couldn’t come up with different slogans. And what happened to the applicants when they couldn’t really change their application letter. The priming effect says that when you use AI to produce a text, it will be much more difficult, cognitively speaking, to come up with new, with your own words, with your own formulations and your own ideas afterwards.

It has been first observed amongst professional translators while they started working with neural machine translation. They claimed that it was much more time and effort consuming to work on those machine translated texts than on texts that had been translated by humans. The priming effect has also been observed amongst students in an experiment. The students were asked to write an essay using GPT-2, GPT family.

And when they reported back on the experiment, on the writing process, many of them said that they had accepted and taken over many passages that were generated by AI because it sounded so good, but actually it was not what they wanted to say. So there is a very concrete risk here that AI not only tells us how to write, but also what to write and possibly what to think. So lesson number one, when you write with AI, don’t trust AI. Now it’s easier said than done.

Lesson Two: The Plumber Experiment

And I will show you why in lesson number two. Let’s do a small experiment together. Imagine that you have a water leak in your house. You need to call a plumber.

So you go through the small ads and you find two of them. Here they are. Which one would you call? Many, many, many people would call plumber number two for the simple reason that number one has too many spelling mistakes in the ad.

And if you think about it for a minute, it’s absolutely not logical. Because what you want your plumber to do is to fix your drain, not to sit in your kitchen and write a poem about how the water level is rising. But our brains are wired since education, because of history, because of society, in a way that we always and constantly associate fluency, lack of spelling mistakes, lack of a foreign accent with competence, all kinds of competence.

ALSO READ:  Facebook CEO Mark Zuckerberg’s Keynote at F8 2017 Conference (Full Transcript)

And this is so heavily rooted in our society, in our culture, in the way we interact with each other, that it has almost become part of our better judgment. It’s like an intuition. It’s like a gut feeling that we have. Now this is turning into a problem because AI does not make spelling mistakes. So we are lulled into believing that because AI writes it so well, it must be true.

So lesson number two: When you write with AI, don’t even trust yourself.

Lesson Three: Trust Language Professionals

So there’s not much left to trust, right? Don’t trust AI, don’t trust yourself. Who can we trust? Well, you can always trust your driving instructor.