It is a mixture of Google and Wikipedia. A new artificial intelligence consisting of a giant database demonstrates creativity never seen before.
It’s the innovation of the year and you have to test it to believe it? ChatGPT, an artificial intelligence that we can ask anything and that has answered everything, takes the form of a chatbot, a conversational agent. A dialog box to which you can ask questions and which will answer them in a completely natural language by searching a gigantic database.
And the results are incredibly realistic and accurate. A few concrete examples, you hand in a topic for a philosophy thesis, an essay, a math problem, and he does the job in seconds. He writes a coherent text from A to Z, answers a problem by explaining the demonstration to you. I suggested to him the topics for last year’s philosophy baccalaureate “Discussion, is it to renounce violence?” and he put me a perfectly correct essay in seconds. The diabolical thing is that if you ask him the same question twice, he writes another thesis on the same subject. Impossible to see that we cheated.
This artificial intelligence demonstrates creativity, such as imagining a scene from your favorite series for example and writing it taking into account the characters and the way of speaking. Or invent an alternative ending to Game of Thrones, or a story for your children that tells the story of a squirrel lost in the forest.
She can also adapt to the level of the person speaking to her, depending on whether she is asked to explain to a 4-year-old child or “how to respond to a child who asks if Santa really exists”. He will adopt the tone you take when talking to a child. He can even make humor, parody, imitate the style of a writer, a singer.
What concrete uses can you imagine for this tool in real life? Google already has something to worry about. The results are infinitely more precise and more complex than on a search engine, where we are often frustrated by the answers. There we have the impression that it is a human being, not a computer, who answers us.
Is there a risk?
And so this kind of tool will be a revolution in a whole lot of trades. Online support services, subscription services, after-sales service. You will be able to chat with these machines, as you do today with a teleconsultant, without even realizing it. Lots of traders will find help there. IT developers who ask to write code or fix it, with deadly precision. Marketing managers asking to develop business strategy help. Even for us journalists. I’m going to interview Tim Cook, what questions should I ask him? He makes me a list of 10 questions, some quite banal, others quite relevant…
This machine has all the same reception limits. In theory he cannot answer illegal or immoral questions like explaining to me how explosives are made” for example. The problem is that some have already found ways around it.
The real danger is erasing the human/machine boundary. Let’s be clear, the answers, as realistic as they are, are just a synthesis of a lot of data that we made him swallow. AI has no understanding of what is being said. The risk is to believe in it a little too much, to become emotionally attached, to trust the machine blindly.