The Washington Post says it has published over 100,000 personalized AI podcasts.
“We have published about over 100,000 personalized podcasts by now.”
Person
Vineet Khosla
1 episodes · 19 quoted claims
Khosla says listeners finish the Post's personalized AI podcasts more often than its regular podcasts.
“The completion rate of these podcasts, it's actually higher than the completion rate of normal podcasts that we publish.”
The Post's personalized podcast tool struggled with articles that used many third-person pronoun references.
“it has a lot of trouble when you have lot of third person references in an article.”
The Washington Post built an internal generative AI tool called Haystacker that lets journalists search through hours of video.
“We have a tool called Haystacker, which can go through hours and hours of videos.”
Haystacker was created for cases where reporters received large amounts of video or data that would take weeks to review manually.
“Somebody gave a reporter a lot of videos, somebody gave a reporter a lot of data and said, Hey, something's going on over here.”
The Washington Post launched Ripple, a project bringing opinion content from sections across America to its audience through partnerships.
“We recently started something called Ripple, so it's Washington Post dot com slash Ripple, where we are going to opinion sections across America”
The Washington Post's website shows a ticker of other news organizations' homepage content at no charge.
“We are letting other news organizations what they're putting on their homepage almost run for free on our site”
The Washington Post uses a large language model hosted on a private instance, which simplified its information security policy.
“We are using a LLM that we are hosting on a private instance.”
The Washington Post labels AI-generated article summaries as AI summaries.
“We still label AI summaries that this is an AI summary.”
Khosla says the Post has not published fully AI-generated video and does not plan to, and would use stronger disclaimers if it did.
“if we put a full AI generated video, which we haven't done so far and we don't plan on, we would put stronger disclaimers.”
At Apple, users followed Siri's voice onto dirt roads when Apple Maps data was wrong, showing Siri and Maps sharing one voice led to excess trust.
“Siri told me to go there, and that's when we realized the Siri voice and Apple voice being the same voice was actually a problem”
Khosla's first job after his master's in AI was building AI models for loans in the mortgage industry.
“my first job in the mortgage industry. We used to do these AI models for loans.”
Khosla says the 2007 mortgage crisis wiped out his industry and ended his first AI job.
“the year being 2007, when the great mortgage crisis and the financial collapse happened, my entire industry got wiped out.”
Siri's founders noticed Khosla's open source contributions and invited him to interview, leading to roughly ten years working on Siri.
“I was contributing some open source code. The founders of Siri saw my code, they invited me to apply for an interview.”
Uber's mapping team set a target of accuracy within ten meters or ten seconds for maps and arrival time estimates.
“We settled on a metric which was ten meters or ten seconds.”
Uber's routing had to balance a marketplace by considering about one hundred drivers per rider and vice versa, unlike consumer navigation.
“for one rider, you would probably reach out to one hundred drivers to see when they can get to them”
Khosla interviewed with owner Jeff Bezos as part of being hired at The Washington Post.
“I interviewed with Washington Post, they had a great team. I interviewed with Jeff Bezos”
Khosla's University of Georgia AI master's program combined language, philosophy, theory of mind, first-order logic and statistical AI.
“What University of Georgia was offering uniquely at that time, and it still does, is it's an interdisciplinary program.”
Khosla says consumer trust used to rest with news mastheads before shifting to online creators.
“The trust of consumers used to be with the mastheads, right? You would read a newspaper because you trusted that there were standards and procedures and professionals”