Prompt assistance

No, you haven’t mastered ChatGPT. Unless you’ve been trained on it. Artificial intelligence wants us to know how to talk to it. But does this involve negligible effort?

By Kevin Erkeletyan

Prompt error. I thought Generative Artificial Intelligence (GenAI) was a long, calm river where you sail along without having to row. Zero out of twenty: review coding “It takes quite an effort,” says Eneric Lopez, director of Microsoft’s National Initiative for AI. His colleague Philippe Beraud, Chief Technology & Security Advisor, goes even further: “It’s essential to contextualise its interactions, and describe and specify what is expected of it, like adapting the response to a particular audience.”

But does this “effort” towards the machine entail consequences? “‘In his In Praise of Software Bugs, Marcello Vitali-Rosati shows that AI, and digital technology in general, imposes a certain work format on those who use it,” says philosophy researcher Luca Paltrinieri. “Office isn’t called Office by chance. It’s a tool for doing bureaucratic work, and this bureaucratic mentality is transmitted through it.When I use Word, I unconsciously learn a certain number of assumptions that I apply to my life as if they were natural.”

THINKING MIDJOURNEY: HALFWAY DOWN THE ROAD?

It’s true: ChatGPT and Midjourney make this conversation “natural”. Midjourney asks you to start each prompt with “/imagine”. For example: “/imagine a group of samurai celebrating a victory in manga style”. However, on the application’s public feed, what was written was the following prompt: “/imagine group of samurai soldiers celebrating victory. Heroic atmosphere. Manga/anime style.” In a sense, it’s a departure from the “natural” structure Midjourney wanted. But the style is telegraphic, almost mechanical. Is this what Philippe Beraud calls “the art of prompting”?

“WHEN COMPUTERS CAN SPEAK HUMAN…”

He points out that “the need to make explicit what is implicit in our way of thinking can be seen as the risk of “formatting” the way we think” – and some may see it as “the acquisition of new skills like algorithmic thinking.” The words are out: algorithmic thinking. This mindset is essential if we are to ”make effective use of AI.” Philippe Beraud defines it as “the ability to solve problems systematically: a valuable skill in many fields, not just mathematics or computing.”

“This does not necessarily mean that we are starting to “think like a machine”. On the contrary, it can help us clarify our own thinking and understand more clearly how we solve problems,” says Philippe Beraud. Nor is this effort a one-way street. He cites Bill Gates: “When computers can speak “human”, every human can speak with computers.” AI is moving towards people – and making the user experience more intuitive – but people shouldn’t just sit back and wait. At Microsoft, Béatrice Matlega’s job is to get them to go towards it, to develop “digital literacy”, and to ensure that young talents “understand how to question an AI.” Luca Paltrinieri agrees: “When you know the codes, you can also change them.” It’s almost a virtuous circle.