Jeff Bezos’s Washington Post Launches an AI Product After Losing 250K Subscribers

The folks implicit astatine the Washington Post person had a unsmooth week, losing much than 250,000 subscribers aft proprietor Jeff Bezos ordered the paper to not marque an endorsement successful the 2024 election. But hey, look implicit there, it’s a shiny caller AI chatbot!

“Ask The Post AI,” arsenic it’s aptly called, is simply a caller chatbot from the Post that returns AI-generated answers to questions based connected the newspaper’s backmost catalog of stories published since 2016. For instance, if you inquire it however galore homes were damaged successful Colorado’s 2021 Marshall Fire, it returns a effect saying, “The Marshall Fire successful Louisville, Colorado, successful December 2021 destroyed much than 1,000 homes, making it the astir destructive blaze successful the state’s history.” That is correct! Below the reply are links to the Post’s related stories that it culled the accusation from.

“Generative AI and the emergence of conversational formats are opportunities for america to gully upon our agelong bequest of fact-based, thoroughly reported journalism to delight and pass readers successful caller ways,” the Post said successful an announcement.

The Washington Post’s caller chatbot provides AI-generated answers to questions based connected the newspaper’s backmost catalog of reporting.

Media companies person been looking for ways to support themselves against the emergence of chatbots similar ChatGPT, which could endanger their businesses if radical usage chatbots to larn astir quality events alternatively than going to quality websites directly. Some outlets person begun partnering with AI companies to person immoderate compensation successful speech for utilizing their enactment arsenic grooming data. Others person alternatively been suing to halt the unauthorized usage of their reporting. And successful the lawsuit of the Post, of course, it’s trying to marque its ain chatbot.

The paper acknowledges that chatbots person a inclination to hallucinate oregon marque up worldly that isn’t true. After all, chatbots are conscionable statistic programs. They conjecture what connection should travel aft the 1 that precedes it but can’t construe information versus fabrication oregon differentiate betwixt morganatic and fishy accusation sources.

That being said, chatbots thin to execute amended erstwhile they’re fixed a constricted acceptable of curated information alternatively than being allowed to propulsion from immoderate root connected the web. “By limiting the hunt results to our published work, we are ensuring that each portion of accusation synthesized by the AI is based connected enactment antecedently published by The Washington Post newsroom.” The Post says that if its AI isn’t capable to find capable capable reporting to make a response, it won’t reply. In regards to vigor usage, the paper says it optimized the programme to beryllium agnostic regarding which AI models it uses, and claims the programme requires nary much vigor than moving immoderate emblematic software.

Funnily, the announcement nonfiction states that the chatbot “may not ever relation precisely arsenic we anticipation — which is wherefore we are asking you to corroborate the results with the published articles.” Seems similar that would decision the purpose.

There’s immoderate statement astir whether AI chatbots would gully postulation distant from autarkic websites. Some reason that radical volition ever privation a deeper read, and volition usage chatbots arsenic specified jumping-off points oregon to reply elemental matter-of-fact questions similar “why is the entity blue.” The Post’s chatbot decidedly could beryllium utile for the former, arsenic hunt connected quality websites is often rather bad. But anyone who has utilized Google precocious knows its AI-generated replies fixed to the apical of searches are bully astatine answering a batch of questions (whether those answers are close is different question). It seems intolerable to contradict that some postulation volition beryllium eaten up by AI.

The Post hopes a conversational chatbot tin service arsenic capable accidental to “delight and pass readers.” The paper hasn’t precisely delighted readers precocious since it was learned that proprietor and Amazon co-founder Jeff Bezos decided astatine the eleventh hr that the paper wouldn’t endorse a president. In airy of Bezos’s caller nationalist enactment for President Trump connected Twitter, and the president-elect’s long-running disdain for Bezos, it looked a batch similar helium was trying to curry favour oregon forestall immoderate further animus. The Post mislaid implicit 250,000 subscribers successful the aftermath of the news—not bully for a paper that already wasn’t doing well. It conscionable goes to amusement that though Bezos has wealth to enactment the Post, being owned by a businessman with different interests comes with its ain problems.