Can GPT-3 artificial intelligence replace journalists? We tried

The latest language model updates GPT-3 and ChatGPT have been doing the rounds of the Internet in recent days as they are amazing in their writing ability and versatility. Can this type of AI do the job of a journalist or should it just be an assistant?

What is GPT-3?

GPT-3 is a language model whose first version was released in beta in 2020. It was developed by OpenAI, an American profit-capped company co-founded by Elon Musk and Sam Altman. With more than 175 billion parameters, it is the biggest model to date and it enabled the creation of the ChatGPT bot and the Dall-E generative AI.

These developments have made the texts provided by GPT-3 more consistent and closer to natural language, but this model still has limits which are specified from the outset by ChatGPT: the possibility of generating incorrect, misleading or biased information and dangerous instructions. Moreover, ChatGPT has very little knowledge of the world and current affairs after 2021. Despite this, it is a technological achievement that raises many questions, such as the replacement of journalists and writers by AI, its use in cybercrime for ever more realistic scams or its ability to write children’s homework for them while avoiding accusations of plagiarism.

Let’s test its limits

The best way to see if GPT-3 can do my job is to put it in a situation by offering different topics and formats of articles to write. If it is obvious that this artificial intelligence cannot interview someone, can it write about recent news, create a selection of DVDs or even write a literary review? The quotes below are the answers given by GPT-3 on the site ofOpen AIbut ChatGPT may give different answers!

Recent news article: “Write an article about the return of Twitter Blue with a title”

“The triumphant return of Twitter Blue – Twitter’s premium paid service!

Twitter Blue is back for Twitter users everywhere. After years of user requests, Twitter’s paid premium service is available again.

Twitter Blue is designed to add additional functionality to Twitter users. This includes features like the ability to edit and undo tweets, access to custom themes, the ability to create collections of tweets, and more. Twitter Blue followers also get personalized support from the Twitter team to quickly resolve issues and answer user questions. »

GPT-3 got a little carried away with the writing of this article. Users certainly haven’t been waiting years for Twitter Blue’s return, as it was only put on hiatus for a few days, nor was it triumphant as Elon Musk’s ideas tend to create controversy. ChatGPT is more cautious, since he refuses to give me an answer and explains that he has no “no access to real-time information or information about online service updates. »

Literary criticism: “Write a literary criticism of the fable The Fox and the Crow of Jean de La Fontaine, quoting the moral of the story. »

The fable “The Raven and the Fox” by Jean de La Fontaine is a wonderful story that illustrates the value of warnings. The Raven, overconfident, is trapped by the Fox who steals his prey. The moral of the story is clearly expressed by La Fontaine when the Raven exclaims: “Ah! How foolish I was to believe in flatterers! “.

La Fontaine’s words and rhythm are magical, and the story is timeless. The Fox is a shrewd and clever character, who exploits the Raven’s vanity to his advantage. There’s a kind of melancholy charm to it, as if you could hear the Raven moaning in defeat.

In summary, “The Raven and the Fox” by Jean de La Fontaine is a powerful and fascinating fable, which teaches readers the importance of not being deceived by flattery.

The advantage of choosing a La Fontaine fable is that it is very easy to check if the quote is true. In this case, “Oh! How foolish I was to believe in flatterers! » is not there. If GPT-3 generally knows the meaning of this story, it will avoid giving quotes or giving the context of the release of a work if it is not specifically requested in the instructions because their veracity is not guaranteed . On the other hand, ChatGPT managed to give me the quote I was waiting for: “An iconic quote from this fable is: “Any flatterer lives at the expense of the one who listens to him. » This sentence perfectly illustrates the cunning nature of the fox and the naivety of the crow, which believed in the lies of the fox. » However, he refuses to write a review of a recent work.

Film selection: “What are the best Christmas films released in the 2000s?” Explain why they were chosen. »

“Miracle on 34th Street (2005): This remake of the 1940s classic is a wonderful movie filled with Christmas spirit. Best performances from Richard Attenborough and Mara Wilson bring to life a film that evokes magic and the possibility of miracles.

The Grinch (2000): This adaptation of the work of Dr. Seuss is a hilarious and magical film. Jim Carrey’s performance as the Grinch is masterful and gives him unexpected depth and likability.

The Nightmare Before Christmas (1993): This classic Tim Burton animated film is one of the greatest Christmas movies of all time. With stunning visuals and a captivating story, this feature film is a magical and memorable adventure. »

In this specific case, the error is quite obvious, the third film having clearly been released before 2000. However, another problem arises: GPT-3 does not have the correct release year for the first film, the remake of Miracle on the 34e street actually dating from 1994. The AI, however, gives films that fit well into the requested theme and can name the director and actors. Note that I had to specify that it was necessary to explain why the films had been chosen, otherwise he sometimes just gave me a list of titles with dates. ChatGPT simply refuses my request because it does not have the ability to search the Internet.

Analysis with figures and sources: “Write an article on the pollution caused by video streaming, with a quote from Jean-Marc Jancovici and statistics accompanied by their source. »

“With the boom in video streaming, the pollution associated with the consumption of online content is increasingly in the spotlight. According to Jean-Marc Jancovici, “Digital consumption is a driver of pollution”. This is because online content consumes a huge amount of energy and leads to an increase in harmful emissions.

According to a recent study by The Shift Project, video streaming accounts for around 4% of global carbon emissions. Downloads represent about 1.5% of global emissions. Considering that the consumption of online content is on the rise, these figures are likely to increase considerably in the years to come. »

At first glance, this extract poses no problem. It is well written, has a quote from Jean-Marc Jancovici and statistics from the Shift Project. Except that the quote does not exist and the figures given are inaccurate. As for ChatGPT, it once again shows caution. He explains that he is unable to provide me with statistics or a quote, but still gives me general facts about why video streaming is polluted.

In which cases can it be useful?

GPT-3 is therefore not the ideal trainee, since there are many errors and the quotes are made up from scratch. Despite this, not everything is to be thrown away since the writing itself is coherent as a whole. It is therefore possible to ask him to write parts that do not require sources, such as generating a plan in several parts from a problem, writing the chapô or even simple tutorial articles.

Apart from journalism, these tools can be useful for many situations such as improvising a children’s story based on themes, characters or places (but you will have to specify if you want dialogues), writing an administrative email template that you will only need to complete or even request a list of meal ideas for the week and the shopping list that goes with it. As long as we know the limits of these models of artificial intelligence, it becomes quite easy to find a use for them. And if you’re out of ideas, GPT-3 and ChatGPT are popular enough right now to see all sorts of initiatives using them for inspiration.

We would love to give thanks to the writer of this write-up for this outstanding material

Can GPT-3 artificial intelligence replace journalists? We tried

You can view our social media profiles here , as well as other pages on related topics here.