Technology

➚ MOVE FAST: Platform-agnostic AI models. Inflection AI just got $1.3 billion from Microsoft, Nvidia, Bill Gates, and other investors, valuing the company at $4 billion. Its strategy isn’t clear, but it appears Inflection is building lots of massive AI models for specific use cases, bucking the exclusive partnership template that Microsoft and OpenAI created with their tie-up.

➘ BREAK THINGS: Platform-exclusive video games. If there’s a takeaway from the fireworks show in federal court between Microsoft and the U.S. Federal Trade Commission, it’s that nobody likes exclusive video game deals: Not the agency, not Sony, not Microsoft’s Satya Nadella, and not Activision CEO Bobby Kotick. Call of Duty for every console!

AI has a lot of great use-cases. Add “talking with the son of God” (according to Christianity) to the list. On Thursday, we watched a Twitch stream of people asking an AI Jesus questions and getting back video responses. Those queries were racking up server costs, which were displayed in the lower right corner of the screen (talking to Jesus was never free).

The answers were pretty good. While many of the questions were silly, some users asked about pretty serious theological topics, which seemed part of a sincere effort to learn something new. It wasn’t clear which model the AI was using, but it was definitely trained on some sort of religious text. We’re not experts on whether this kind of thing is blasphemy.

Using AI to imagine what celebrities, fictional characters, or even religious deities might say in response to questions is one of the fun uses of AI. Character.AI, which specializes in this area, has seen massive growth in a short period of time. We counted more than 30 versions of Jesus, including a “drunk” one (too much water into wine?) and an “emo” one on the platform.

The 26-year-old Washington AI whisperer

THE SCENE

While OpenAI’s Sam Altman has been dominating the spotlight, another artificial intelligence CEO is increasingly influencing how the industry is perceived by Washington regulators.

Alexandr Wang, the 26-year-old chief executive of the billion-dollar company Scale AI, which manages an army of human AI trainers, was a familiar D.C. face long before the craze over ChatGPT began, securing lucrative government contracts and winning over members of Congress.

Wang, who dropped out of the Massachusetts Institute of Technology to found Scale in 2016, has taken particular interest in speaking to lawmakers about his hawkish views on China, which he says is the greatest geopolitical competitor to the United States.

“China has stated that it will dominate AI in both the public and private sectors by 2030 and is throwing the resources behind the goal to make that happen,” Wang warned at Scale’s inaugural tech summit in the capital earlier this month.

What really makes him stand out in the California-centric tech industry, however, are his deep connections in Washington. “Alex and I are actually friends, not like fake DC friends, like real friends — we hang out,” Rep. Mike Gallagher, R-Wis., said during the summit. “We’ve known each other for three years now.”

In recent months, the CEO has twice briefed the new Select Committee on the Chinese Communist Party that Gallagher chairs. “I found him to be very insightful, and at the same time personable, and just energetic about the issues of AI,” said Rep. Raja Krishnamoorthi D-Ill., the senior Democrat on the committee who met Wang during a recent trip to California.

Other lawmakers had similarly complimentary things to say about Wang, whose company spent over $1 million on federal lobbying last year, according to public disclosures. “He strikes me as a man on a mission — fiercely patriotic and fiercely determined to ensure the U.S. achieves AI dominance,” Rep. Ritchie Torres, D-N.Y., told Semafor.

A spokesperson for Scale declined to make Wang available for an interview. Vijay Karunamurthy, the company’s field chief technology officer, told Semafor that Scale has been engaging with the U.S. government about AI for years, unlike firms who only woke up to the technology recently.

He said Wang has coined a term for them: “AI tourists.”

LOUISE’S VIEW

Scale’s outsized presence in Washington may open lawmakers’ eyes to the immense amount of human work that goes into developing AI. ChatGPT was trained on billions of text snippets from the internet, but what makes its answers sound lifelike is a process called reinforcement learning from human feedback (RLHF), essentially having a real person rate the AI’s responses and correct them for inaccuracies.

Scale and its competitors are the ones responsible for providing that labor, as well as ensuring it gets recognized. “We absolutely think it’s important to acknowledge the role that human feedback and human alignment plays in making these models actually useful for different use cases,” Karunamurthy said. “It’s not just understanding how the sausage is made, it’s actually the really critical piece to make AI useful to human beings.”

To read the Room for Disagreement and the rest of the story, click here.

Number of people who have signed up for Snapchat+, the messaging app’s $3.99/month subscription service. That’s slightly less than half the amount who subscribe to The New York Times Company’s various properties. Snapchat+ gives users access to a host of largely cosmetic upgrades, and they can also see who has rewatched their story.

The U.S. Federal Trade Commission published a paper Thursday warning of potential consolidation and anti-competitive behavior in the generative AI industry. The agency specifically called out cloud providers without naming them. It’s clear the FTC is talking about Microsoft, Google, Amazon, and Oracle, all of whom are offering services to AI startups and big, established companies.

As we reported in Wednesday’s scoop, there’s competitive price pressure on cloud companies at the moment. The FTC accurately pointed out that the open source market for AI models could disrupt those cloud providers and ramp up competition.

But it seems unlikely that the watchdog will bring any lawsuits in the generative AI industry any time soon. Still, the paper could be read as a shot across the bow of big tech companies who have been known to stymie innovation with monopolistic tactics. For a change, Lina Khan’s FTC is thinking ahead and potentially preventing anticompetitive behavior, rather than waiting until it’s too late and bringing lawsuits that take too long to have much of an impact.

In another development related to the FTC lawsuit against Microsoft’s Activision deal, lawyers for Sony gave away important company secrets when they used a black marker to redact documents submitted in the case.

Sony, the maker of Playstation, opposes the deal, and the redactions, which didn’t really cover up the text, tell us why: There are a million Playstation owners who do nothing but play Call of Duty, which is owned by Activision. Sony worries Microsoft might make the game exclusive to Xbox.

And about those redactions: Sony’s lawyers obviously should have been using a computer instead of a marker to block out confidential information. But would you trust AI to do it instead? I asked the founder of AI e-discovery startup Logikcull.

LinkedIn is going after fake accounts by implementing a new AI-generated image detection technique. That’s great news, as even real people can be incredibly spammy on LinkedIn. Nobody needs bots in their inbox, too.

Semafor would like to ask, though, whether LinkedIn could make an exception for using AI to touch up photos. Research has shown that humans are biased toward attractiveness (usually measured by facial symmetry). Perhaps AI could help equalize the playing field.

There are many new tools that can make professional-looking headshots for people that stay true to the way they actually look. The author of this post was called handsome by the omniscient Kara Swisher and would never use one of those services. But for people who don’t have the means to get professional headshots and some light photoshopping, they could be a great resource.

— Reed

• What a Chinese heat wave means for the world.
• OpenAI is training AI models on Microsoft’s cloud services, Reed Albergotti scooped. The cloud computing industry is expected to grow north of $2 trillion by 2030 and the fight for market share is intensifying.
• Prigozhin may have colluded with other Russian officials before his failed coup. Read our curated analysis from experts.