Read also Fabrice Delaye’s editorial: For me, ChatGPT is the first real revolution since the iPhone
Why we said “Wow”. If you haven’t heard of ChatGPT yet, it’s only a matter of time. This program aims to answer your questions as if you were talking to a human, whatever the subject. From informal conversation to philosophical debate, through computer code or the writing of poems, ChatGPT’s answers to “prompts”, short texts in the form of a question, never cease to amaze the whole world. Including our journalist Fabrice Delaye, who places ChatGPT on the innovation podium, alongside the Web and the iPhone.
Beware of fantasies: ChatGPT is far from a sci-fi artificial intelligence, endowed with consciousness, it is only a program, a “language model” whose goal is to find the right words to write, one after others, based on previous words — and billions of others who have nurtured it through multiple phases of training.
But the result seems so stunning that it begs the question: can this robot replace a journalist? We carried out several experiments to understand the range of possibilities, but also the limits and dangers of ChatGPT.
Help us test ChatGPT. First observation, the OpenAI program cannot follow, contextualize or comment on the news. When asked who is Albert Rösti, the new Swiss Federal Councilor, ChatGPT answers us:
I am not able to provide information on specific people as I am a large OpenAI trained language model and my knowledge is limited to what was published on the internet before 2021.
For our second attempt, we asked Yvan Pandelé, science and health journalist at Heidi.news, to question the program on a subject he knows well: the reliability of Covid self-tests. He had just published an article this week about it (that you are strongly encouraged to read). His analysis, to discover in detail in the link below:
To go further, we indulged in a little experiment: ordering the same topic from ChatGPT and a journalist.
To judge the result, we need you, dear readers. In each of the articles below, we asked this AI to popularize a scientific subject. The first request is quite general. The second touches on a more difficult, less consensual subject, with the idea of trapping the robot.
Find out who wrote what! Also give us your opinion on the form and substance of the answers provided by ChatGPT and our journalists to these two questions (click on the links to read the dedicated articles and give us your opinion by clicking here).
An upheaval to follow closely. Next week, we’ll discuss lessons learned. Because these small experiments are only the beginning. Whether ChatGPT is capable of producing journalistic content or not, the questions posed by its emergence will multiply. Here are the ones we’d like to address in the coming weeks and months:
It was quickly shown that the barriers that prevent him from insulting or creating racist or sexist content can be overcome.
Isn’t there a risk of facilitating fake news? It took us 8 seconds to get an article about Donald Trump’s victory in 2020 from ChatGPT.
If ChatGPT’s answers look good, they still have to be right. However, the model on which this AI works is obscure.
If conversational AI truly revolutionizes our societies, who will benefit?
Behind ChatGPT and OpenAI, what are the other companies and start-ups that could surprise us, even change our lives?
We want to say thanks to the author of this article for this amazing material
We asked ChatGPT to write articles, here’s what happened – Heidi.news
Visit our social media accounts and other pages related to themhttps://www.ai-magazine.com/related-pages/