An AI has written almost 80 articles for the US website CNET since November. The problem: Despite subsequent fact checking by a human editor, some of the texts contain gross errors.
CNET is one of the largest tech websites in the United States. In addition to classic tech topics, the website also deals with areas such as wellness, culture, science and finance. In the case of the latter, an experiment was dared at the end of last year: an AI was supposed to write some of the articles.
The focus was on topics that required little creative work and had a simple goal: to explain financial terms. The first of these AI-based articles appeared with the title What is a Credit Card Charge-Off
am 11. November 2022.
An AI as author – the network is appalled
The artificial author was brought to light by SEO expert Gael Breton. On Twitter, he pointed out that Google-optimized content at CNET is now also being written by an AI. The articles first appeared under the author’s name CNET Money Staff, later only under CNET Money. If you go to the author’s page, you will find the short information there:
This article was powered by an AI engine and reviewed, fact-checked, and edited by our editors.
There was no official announcement from CNET that they wanted to use AI as an author. Suspected in connection with the hidden information about the AI author Futurismthat they wanted to keep the project out of the public eye for the time being.
link to Twitter content
Rightly so, as it turned out. Because after it became known, the artificial author from CNET caused a loud, mostly negative response. So labeled Luke Plunkett from Kotaku considered the whole thing Awful
Mike Murphy von IBM Research summed up the process by saying: Content by robots (text generator) for robots (SEO crawler).
Another frequently mentioned concern: In the future, such an AI author could cause problems, especially for many young professionals and young authors. Because it is precisely their texts that an AI would write faster and cheaper. In the long term, however, the industry would primarily harm itself.
CNET responded with a post on their own website, but did little to address the concerns expressed. Above all, they explained their own thoughts on the experiment and praised the transparency that they had shown. If successful, the AI could relieve authors in particular and enable them to focus on creative articles. In addition, more and better articles may be possible for the website.
This AI cannot yet claim to be better than a real author. Sören’s self-test shows that artificial intelligence lacks empathy.
more on the subject
I had ChatGPT rewrite my book
CNET’s AI articles contain blunders
However, the experiment has already suffered at least one severe setback. Because the articles that CNET had written by an AI contained serious errors, despite human editing. Futurism In an examination of the texts, he found several places where the AI was technically wrong.
In one of the examples given, the AI claims:
For mortgages, car loans, and personal loans, interest is usually calculated in simple numbers.
For example, if you take out a $25,000 car loan and your interest rate is 4 percent, you pay $1,000 in interest per year.
This calculation is incorrect. Because the interest rate is always only added to the remaining unpaid amount. Once part of the $25,000 has been paid, the interest amount will be less than $1,000 and will decrease with each subsequent transfer.
In the article, Futurism lists two other examples where the AI was also fundamentally wrong in its statements.
CNET has since responded. The articles written by the AI are still available, but are now introduced with an info box. In this, CNET points out that the article is currently being checked for accuracy. If errors are found, they will be corrected and pointed out in the article.
Meanwhile, artificial intelligence is trying not only as an author, but also as a lawyer. Such robot lawyers are already a reality in the USA.
more on the subject
AI as a lawyer in court: Soon also in Europey?
How do you feel about this? Do you believe in the potential of AI as an author? Or should articles be written by people for people? Write us your opinion in the comments!
