Michel C. Auger: I was curious. I wanted to see what ChatGPT could say about me. What was my surprise…
“Michel C. Auger was a Quebec journalist well known for his work as a political columnist and investigative journalist. He was born on July 21, 1950 in Montreal… Over the years, he notably covered the case of the Charbonneau commission, an investigation into corruption in the construction industry in Quebec.
Unfortunately, Michel C. Auger was murdered on September 8, 2000 by a gunman as he was leaving the CKAC studios…”
I’m used to confusion. There are two journalists named Michel Auger. My eldest was one of the greatest reporters to cover organized crime in Canada. As he was older than me, when I started in journalism, I immediately added my initial to my signature precisely to avoid any confusion. But, above all, we became friends.
So I’m not surprised that ChatGPT confused us. But the amalgamation of our respective biographies contains so many errors that we would laugh if it weren’t a little sad to see what artificial intelligence is doing to us.
First of all, I wasn’t born in 1950, but neither was Michel (1947). I was born in Montreal, but he was in Shawinigan. It starts badly, let’s say…
Alexander Sirois: You murdered! We swim in delirium.
As far as I know, you are the first person to have had such an experience. ChatGPT does not provide obituaries to living people on a regular basis. Fortunately. Nevertheless, his hallucinations are frequent. We asked him about the members of our editorial team. If there is sometimes some truth in his answers, many things have been invented from scratch.
We thus learned that our chief editorial writer, Stéphanie Grammond, worked on The Press “until his retirement in 2020”. For my part, according to ChatGPT, I currently work at Duty. Philippe Mercure was a parliamentary correspondent in Ottawa for RDI. Nathalie Collard is the author of several books on arts and culture, including dance with life And The body on show. Vincent Brousseau-Pouliot “worked for Radio-Canada and was also a foreign correspondent for various Canadian media”.
As for Serge Chapleau, he would have created “the fictional character the little guy from Shawinigan, which he uses to comment on political and social news”.
This is all wrong.
Michel C. Auger: What worries me the most is the lack of intelligence of artificial intelligence. We had to get to the point where the AI would be able to verify its information or, at least, put facts in relation to each other.
But, according to ChatGPT, I would have covered (or is it Michel?) the Charbonneau commission. I didn’t cover it and, to my knowledge, neither did Michel since he was already retired.
But even more disturbing, since I died in the year 2000, how could I have covered a commission that began its work in… 2011! It’s not just a mistake, it’s showing that AI still has major factual gaps. We are not talking about analysis here.
Alexander Sirois: I don’t know if this can reassure you, but I just asked ChatGPT-4 about you. I am talking here about the new version of the conversational robot (developed by the Californian company OpenAI). The experts all say that the rate at which these robots are improving is exponential. Result: he knows you’re not dead!
That’s the good news.
The bad news is that ChatGPT-4 still talks a lot of nonsense. You would be, according to him, a journalist “specialized in the coverage of legal affairs and national security”. And you would have already “worked for the TVA network” where you would have “hosted the public affairs program Behind the scenes of power “.
In short, there is progress, but this robot still has crusts to eat…
Michel C. Auger: Thank you for resuscitating me. But I want to come back to Michel’s death. (“Without C”, he used to say). Because there, artificial intelligence shows great dispositions for fiction. Not only was he not murdered, but the attempt on his life – which did take place – was not as he left the CKAC studios, but in the parking lot of the Log. By the way, we both worked for several years at the Montreal Journalwhich is nowhere mentioned!
But Chat GPT adds a layer: “The investigation into his death lasted several years and several people have been charged in connection with this case…” But no, no one has been charged. The perpetrators of the attacks were never found – in fact, it should be said that they were never formally charged.
If there’s anything that really bothers me about all of this, it’s not that a few facts may have been wrong. As journalists, we’ve all made mistakes, so we can’t throw stones too quickly. But when the AI doesn’t have the facts, it makes them up. That really bothers me.
Alexander Sirois: THE wall street journal recently revealed that two Google researchers tried to convince the leaders of this company, in 2021, to release a ChatGPT-type conversational robot. They were told no. In particular because it was considered, at Google, that the program “did not meet the standards of security and fairness of artificial intelligence systems”. But last November, the OpenAI company opened Pandora’s box by releasing ChatGPT. Since then, we have witnessed a frantic race between the digital giants. It looks like they are now playing who can piss the farthest.
And we’re starting to pay for it. This is among other things why a series of experts (including Montrealer Yoshua Bengio) have just asked the companies that design these systems to put their work on hold for at least six months.
And on Friday, Italy blocked the use of ChatGPT over concerns over the use of user data.
These artificial intelligence tools have many qualities. Tremendous potential, too. But they are not ready yet. Your false death is a clear demonstration of this!