Om automatiske nyhetstjenester og robotene som leser dem

Automatiske nyhetstjenester i kombinasjon med generell menneskelig sløvhet kan under optimale forhold rote det skikkelig til. Tidligere denne uken gikk aksjekursen til det amerikanske flyselskapet United Airlines rett til bunns etter at en kjede av tilfeldigheter fikk en flere år gammel nyhetssak om mulig konkurs til å boble opp til overflaten og havne som toppsak hos Google News og Bloomberg:

"The comedy of errors began with just one reader who went to the South Florida Sun Sentinel’s website and viewed a 2002 article on United Airlines’ bankruptcy.

That single visit in the early hours of Sunday morning, a period of low traffic, apparently bumped it into a "Popular Stories" in the business section.

At 1:37am, an electronic Google program swept through the paper’s website for new stories and spotted the link.

Google says its program scanned the piece and, seeing there was no 2002 dateline, indexed the article for inclusion on its news pages.

Three minutes and two seconds later, Google News readers started viewing the story on the Sun Sentinel’s Web site.

A Florida investment firm found the story on Monday morning with a Google search and posted a summary on the Bloomberg financial information service.

That visibility – Bloomberg is seen by thousands of investment managers and traders – sparked the run on United shares
[…]
Investors then dumped the stock at a huge rate and here algorithms again played their part.

Experts said the automated trading programs were applied to the trading of shares based on market-moving information trawled from the internet.» Les hele saken hos times.online