Bloomberg Has a Rocky Start With A.I. Summaries

Bloomberg, the Financial News power, has experienced the use of artificial intelligence to help produce his journalism.

It has not always gone without hitches.

The news outlet had to correct at least three dozen summaries generated by articles published this year. One happened on Wednesday, when Bloomberg broke news On the car rates of President Trump.

The article correctly reported that Mr. Trump would announce the rates as soon as that day. But the summary of the projectile point of the article written by Ai said inaccurately when a wider tariff action would take place.

Bloomberg is not only in trying the IA – many stores are understanding the best way to embrace the new technology and use it in their reporting and editing. The chain of Gannett newspapers use Similar synthesis generated by the I on its articles and the Washington Post has a tool Called “Ask the post” that generates answers to the questions of the published articles.

And the problems have sprouted elsewhere. At the beginning of this month, the Los Angeles Times removed its tool from an opinion article after technology described the Ku Klux Klan as something different from a racist organization.

Bloomberg News declared in a declaration that publishes thousands of articles every day and “currently 99 percent of the summary of the AI ​​satisfies our editorial standards”.

“We are transparent when the stories are updated or correct and when the IA has been used,” said a spokesperson. “Journalists have full control over the fact that a summary appears – both before and after publication – and can remove it that they do not satisfy our standards.”

The summaries of the AI ​​are “intended to integrate our journalism, not to replace it”, added the declaration.

Bloomberg announced On January 15, which would have launched the summaries generated by the AI ​​at the top of the news items. The summaries consist of three points list that condense the main points of the article.

John Michlethwait, Bloomberg’s editor in charge, established the thought on the summary of the AI ​​in a January 10th wiseWhich was an extract of a lesson he had kept in City St. George’s, University of London.

“Customers like: they can quickly see what it deals with. Any story. Journalists are more suspicious,” he wrote. “Journalists fear that people simply read the summary rather than their history.”

But, he recognized, “a summary of the AI ​​is as good only as the story on which it is based. And obtaining the stories is where humans are still counted”.

A summary was removed from a March 6th item Because he said in an incorrect way that Mr. Trump had imposed rates for Canadian goods last year instead of this year. Another, March 18th item Information on the managers of sustainable funds, “he has not managed to distinguish between active and passively managed funds, consequently providing incorrect figures”, according to a correction.

Other errors have included incorrect data, incorrect attribution and references to the wrong American presidential elections.

For Wednesday scoop During the tariff announcement of Mr. Trump, a correction was soon added which noticed that the summary had been removed because it was “erroneously eliminated when the wider tariff action occurred”. An updated version of the correction has attributed it to the lack of “attribution on tariff times”.

Bloomberg’s spokesman said that the feedback has been positive for summaries in general “and we continue to perfect the experience”.