Technology

➚ MOVE FAST: Pays for itself. South Korea approved more than 60 new R&D projects, including an AI accelerator for advanced chips to take on Nvidia. It’s part of its high tech push, spending $1.5 billion more this year. Korean firms like Samsung have already benefited from the AI boom, boosting profitability.

➘ BREAK THINGS: Pay to play. Newly unsealed court documents show Apple received $20 billion from Google in 2022 to be the default search engine in Safari. To put it in perspective, the agreement was worth twice Airbnb’s annual revenue last year and accounted for 17.5% of Apple’s operating income in 2020.

Yesterday, I spoke with Adam Jackson, CEO of blockchain-enabled talent network Braintrust, about the company’s new AI recruiter tool, called AIR. The large language model-based system reads through thousands of resumes and then conducts automated spoken interviews with candidates, distilling the pool and delivering the best prospective hires to clients.

The product, launched Monday, has created a debate about the future of human job recruiters. Jackson, a serial entrepreneur and former CEO of Doctor on Demand, told me he’s had to calm employees down, telling them the commotion is a sign they’ve created something impactful.

But I think the real debate about AIR should be about something else: bias. Jackson claims the tool can already outperform human recruiters on limiting unconscious bias because it’s instructed to ignore all information pertaining to race and gender.

Experts in this area are skeptical. I spoke recently about this general topic with a prominent plaintiffs lawyer, who said software companies claiming to employ AI to reduce hiring bias give a false sense of confidence to their clients.

Some algorithmic hiring tools perpetuate bias because they base their decisions on datasets that are infused with human predispositions.

Jackson, though, believes the only way to change the status quo is to innovate. And he says AIR differs from older tools because it does not use past performance of job candidates as a rating metric. Rather, it’s a sort of blank slate: The large language model is instructed to look for things that make for good job candidates, but it isn’t picking up on subtle patterns in past data that could perpetuate bias. This idea wasn’t possible before large language models improved in capability over the last couple of years.

The company’s clients include Goldman Sachs, Google and Airbnb. Jackson hopes those firms will keep tabs on how well AIR does on diversity. “I want to know the truth, and if we’re ten degrees off, I want to fix it,” he said.

Travis Kalanick’s new LA mission

THE SCOOP

Ex-Uber CEO Travis Kalanick has been quietly lobbying regulators to force delivery apps like Uber Eats and DoorDash to hand over valuable customer data to restaurants, which would also benefit his food company.

That push has spurred some independent restaurant owners — a group that has often had a strained relationship with companies like Uber Eats and DoorDash — into the arms of the delivery apps, which are vehemently fighting Kalanick’s efforts.

He’s making the move, which pits him against the company that made him famous (and infamous), to help CloudKitchens, a company facilitating food preparation for delivery-only restaurant brands that have no brick and mortar presence.

CloudKitchens’ parent firm, City Storage Systems, also owns software company Otter, which allows restaurants to accept deliveries from multiple services in one app.

Kalanick made his fortune at Uber by ignoring regulators, whom he cast as corrupt taxi industry pawns. His latest push focuses on Los Angeles, according to lobbying records filed in April. The Digital Restaurant Association, which is linked to Kalanick, according to people familiar with the matter, plans to advocate for “regulations on third party online food order and delivery,” which matches the description of similar lobbying efforts in other places like Florida’s Miami-Dade County and in Georgia. A statewide bill in California is also moving through the legislature but has lost the data provisions that originated the effort.

“Restaurants in Los Angeles urgently need your help to defend restaurants from harmful business practices of big tech food delivery apps,” reads a website set up by DRA, which describes itself as “a coalition of restaurant owners and operators, associations, civic groups and concerned citizens.”

The DRA was created by lobbying firm Tusk Holdings, led by Bradley Tusk, an Uber investor and a long time Kalanick associate. It’s pushed a handful of similar bills, none of which have yet been successful, elsewhere in the country that would force food delivery firms to provide data to restaurants that utilize their services.

Some restaurant owners like Leland Neal, CEO of the Taco Rio chain in South Florida, oppose the efforts. “I don’t want it. I don’t need it,” Neal said. “And if I want to get hold of any customers, I just put a coupon on there. We do it all the time.”

Joe Reinstein, the DRA’s executive director and CEO, said restaurants are squarely on their side and declined to confirm whether Kalanick has any connection to the group. “We have had 800 different restaurants sign a petition in Los Angeles County and we have thousands of members nationally. That’s all we are — a collection of independent mom and pops,” he said.

Chamber of Progress, a tech industry coalition that includes food delivery apps, believes the effort is disingenuous. “It’s surprising to me to see how far they’ve gotten given the lack of transparency,” said Chris MacKenzie, a spokesman for the group. “We’ve noted in our communications with lawmakers that this is a Travis Kalanick project.”

Spokespeople for Kalanick, CloudKitchens, City Storage Systems, and Tusk Holdings did not immediately respond to requests for comment.

Welcome to the “zombie internet,” a term coined by 404 Media to describe what’s become of Facebook now that it’s plagued with bots and humans interacting with AI-generated content.

There are a lot of bizarre AI-generated images that go viral, like “shrimp Jesus” or “Uncle Sam Amputee Jesus and a soldier with machine gun” with what appear to be fake accounts leaving repetitive comments such as “Amen 🙏🙏🙏”. These bots drive up traffic, directing Facebook’s algorithms to amplify that kind of content on people’s feeds, which attracts the attention of real users, creating a vicious cycle that increases engagement.

AI spam is making the internet worse, and it’ll only go downhill from here as it becomes increasingly difficult to tell what’s real from fake. Content is becoming cheaper and easier to churn, and people are figuring out how to game algorithms using SEO tricks or fake accounts.

Companies like OpenAI and Microsoft do label AI-generated images, but their efforts to help people identify them don’t seem to be all that effective in tackling the problem.

There’s more AI spam than ever polluting the internet, and the strange mishmash of bots and humans maximizing fake content on Meta is a cautionary tale of what could happen to other sites if nothing is done to clear up the trash.

— Katyanna Quach

Ryan Pannell is CEO and Global Chair of Kaiju Worldwide, which employs predictive AI in investing and tech research. (Note to JPMorgan CEO Jamie Dimon: Let us know if you would like to comment in a One Good Response.)

• Donald Trump finally gets the convention of his dreams.
• FTC plans to recommend a possible criminal case against ex-Pioneer CEO.
• Republicans aren’t worried about saving Mike Johnson.