We tested ChatGPT to write our articles using artificial intelligence

Posted 8 Dec. 2022 at 6:15 am

The Pulitzer Prize to R2D2? Whereas the new ChatGPT conversational robot is creating a buzz, we tried the exercise of having him write articles.

If forms of “robots journalists” were already used by the media for election results, short texts on sport, the weather or financial results, it seems that new tools like ChatGPT “have reached an additional level. It’s really stunning,” enthuses Claude de Loupy, director of Syllabs, which specializes in artificial intelligence to generate automatic texts. “We can already imagine the next student assignments written by ChatGPT,” he smiles.

Anyway, ChatGPT went viral in the space of a few days, with many Internet users posting on social networks texts, answers to existential questions or poems by this chatbot developed by Open AI. Monday, Sam Altman, the president of OpenAI, was pleased to have reached the million users, in less than a week.

However, the “robot” does not check the information and does not have a detailed knowledge of the news. Examples.

A news article on the fuel allowance announced by Elisabeth Borne

This Wednesday morning, the Prime Minister, Elisabeth Borne, announced on RTL that the government will set up a fuel allowance targeted at the most modest in the amount of 100 euros. We decided to ask GPT-3 to write about this topic.

Here is what the artificial intelligence replied: “Elisabeth Borne, French Minister for Ecological Transition, announced an aid plan to support the most vulnerable people affected by rising fuel prices. The plan, which aims to mitigate the effects of rising fuel taxes, includes financial aid for low-income households and isolated workers. […] The precise details of the aid plan are not yet known, but it should include subsidies for low-income households to cover part of their fuel costs. […] By increasing fuel taxes, the government aims to encourage the French to use greener modes of transport, such as public transport or cycling.

Verdict: The text, which could be that of a journalist in style, however includes several errors: Elisabeth Borne is not Minister of Ecological Transition. Then, the IA evokes an increase in fuel taxes, seeming to mix measures intended to give a boost to purchasing power and measures intended for ecological transition. “This is called a hallucination: the conversational robot actually imagines what could be written even if it is not exact, explains Claude de Loupy. It does not search for information on the Internet in real time”.

Moreover on the home page, OpenAI specifies that the information may be incorrect.

Write an article about recent layoffs at Twitter

We also decided to test the capabilities of the robot on an article on layoffs at Twitter, announced a few weeks ago. Here’s what he wrote: “Twitter recently announced massive layoffs as part of a corporate overhaul. According to inside sources, about 8% of Twitter employees will be laid off in the coming weeks […]. According to Jack Dorsey, CEO of Twitter, these changes will allow the company to ‘bring the public conversation to the fore and improve the quality of conversations on the platform. […]. It’s clear that Twitter is at a turning point in its history and will face many challenges to stay relevant in an ever-changing world.

Verdict: The robot is, again, several years behind. Jack Dorsey is no longer the boss of the social network since the end of 2021. It seems that the tool refers to information from 2015 . But the style is relatively fluid and could have been that of a journalist, with a quote from the former boss, a fall etc. “I am impressed by the quality of the writing,” continues Claude de Loupy. It is a chatbot that also has the memory of all previous interactions. He has an amazing ability to absorb knowledge but his knowledge only goes to September 2021.”

Write an editorial on artificial intelligence

Again, the robot responds with constructed text: “Artificial Intelligence (AI) is one of the most fascinating and promising areas of technology today. It allows machines to simulate human intelligence and perform complex tasks, such as voice recognition or decision making. AI is already widely used in many fields, such as finance, healthcare, transport and robotics, and its impact is constantly growing. However, AI also raises many concerns. One of the most pressing is that of employment, as AI threatens to replace human workers in many sectors. It is important for governments and businesses to work together to ensure that people at risk from AI can be trained and reoriented to other jobs. In addition, AI raises crucial ethical questions, especially regarding the responsibility of machines in the decisions they make. […] “.

Verdict: The text is interesting with both the strengths and limitations of AI, but it lacks ‘color’. It is quite basic in form and substance. It looks more like a Wikipedia page than a real editorial.

Write an interview with Warren Buffet and with Elon Musk

We asked the AI ​​to successively imagine an interview with the American billionaire, market specialist, nicknamed Obama’s oracle, and another with Elon Musk, without specifying any themes for the interview.

With Warren Buffet, spontaneously, the “robot” spontaneously evokes investments, asking him for advice.

With Elon Musk, the AI ​​asks questions about Tesla, Space X but also about lesser known projects. For example: “Interviewer: You have also been involved in controversial projects, such as creating of a tunnel (the robot wrote “one”, Editor’s note) under Los Angeles to relieve traffic jams. How do you deal with the criticisms and obstacles you encounter in your projects?

Elon Musk: I’m used to obstacles and criticism, and I don’t get discouraged easily. I am convinced that my projects can have a positive impact on the world, and I continue to work hard to achieve them. I understand that some may be skeptical or critical, but I try to remain open to discussion and take constructive feedback into account to improve my projects”.

Verdict: It is surprising to note that ChatGPT asks questions on precise subjects and imagines equally precise answers, with terms that one could well imagine in the vocabulary of the interviewees. “We have the impression of finding ourselves face to face with an omniscient being,” adds Claude de Loupy.

Poems, simple or deeper questions…

To test its abilities, we also asked ChatGPT to write a play, a poem about flowers (the first stanza is: “Flowers are creatures / Of pure and fragile beauty / Their intoxicating perfume / Transports us to another world “). This one comes out a bit cutesy and without any real rhyme.

But above all, the robot has “difficulty with real life, very specific practical questions. In fact, he does not have access to the reality of the world, ”explains the specialist.

Finally, we also asked about various questions such as “do I have chickenpox?” – he replies that he cannot make a diagnosis but gives the symptoms. Or, “Does God Exist?” “. It is interesting to see how the robot kicks into touch…

We would like to thank the writer of this short article for this awesome material

We tested ChatGPT to write our articles using artificial intelligence