Hello and welcome to Eye on AI.
Leading AI companies OpenAI and Google took turns making headlines on Monday and Tuesday, announcing a slew of new generative AI products and model updates. On Wednesday, the U.S. government had its moment.
A bipartisan group of senators, led by Chuck Schumer (D-N.Y.), released a long-awaited “roadmap” for regulating AI. The 20-page report is the culmination of an almost year-long listening tour during which the Senate AI Working Group held closed-door meetings with more than 170 tech industry leaders, academic researchers, civil rights leaders, and more to better understand how to regulate the technology. The lawmakers called for the U.S. government to start spending at least $32 billion annually on AI “as soon as possible”—and that’s not including what it intends to spend on AI for defense. But while they acknowledged the variety of harmful applications and currently unfolding consequences of AI, they didn’t propose any specific regulation. Instead, they punted it to the Senate subcommittees, laying out areas where they should focus their efforts, including AI training for the private workforce, mitigating AI’s relentless energy demands, and the problem of AI being used to create election disinformation and child sexual abuse material (CSAM).
Many groups, watchdogs, and experts who provided insights during the listening sessions are calling the results a disappointment, lacking vision, a delay of much-needed regulation, and a missed opportunity to act when much of the world already has.
“The nine ‘insight forums’ functioned as a stalling tactic,” said Amba Kak and Sarah Myers West, co-executive directors of AI Now Institute, an AI policy research group, arguing in a statement that momentum to regulate AI was instead diverted into a closed-door, industry dominated—and now industry-benefiting—process.
Alondra Nelson, a former acting director of the White House Office of Science and Technology Policy who participated in the listening forum on supporting U.S. innovation in AI, described the roadmap to Fast Company as “too flimsy to protect our values” and lacking “urgency and seriousness.” Suresh Venkatasubramanian, another former White House official who coauthored the Biden administration’s Blueprint for an AI Bill of Rights, a set of principles intended to guide the development and use of AI systems, said he feels “betrayed” after he and other AI ethicists participated in the discussions “in good faith,” despite concerns about industry regulatory capture.
If there’s one thing Schumer made clear, it’s that U.S. domination of AI is the goal, referring during the press conference to the $32 billion as “surge emergency funding to cement America’s dominance in AI,” including “outcompeting China.” The money would go to R&D, infrastructure, funding the outstanding CHIPS and Science Act, a series of “AI Grand Challenge” programs, and dozens of other government agencies and AI-related initiatives.
Delaying regulation certainly helps U.S. tech companies—especially dominant players, which argue that regulation would “stifle innovation”—to do so. The AI Working Group, according to its own objective as stated in the report, was created because AI is too “broad” and “does not neatly fall into the jurisdiction of any single committee.” It seems all this special group decided, however, is that the subcommittees should be the ones to propose legislation after all and that while billions in taxpayer funding can’t wait, regulation can. Government funding for science and technology is important, and it’s how we got the Internet in the first place. But prioritizing dominance over safety is a dangerous path.
Plenty of AI bills have been introduced in subcommittees only to stall. The Senate Rules Committee did yesterday pass three bills intended to safeguard elections from deceptive AI, but they still need to advance in the House and pass in the full Senate—and the clock until the election is ticking. If the AI Working Group hadn’t taken a year only to suggest zero tangible proposals for regulations, progress on mitigating this clear and present danger of AI could be much further along. In the report, the lawmakers also expressed explicit support for a national data privacy law, which again, is the type of thing they are in charge of introducing and passing.
Schumer told the New York Times, “It’s very hard to do regulations because AI is changing too quickly. We didn’t want to rush this.” Pouring $32 billion a year—plus however much more in defense funding—into AI “as soon as possible” certainly seems like decisive, swift action. At the same time, it’s been almost two years since the launch of a new era of generative AI made the need for regulations ever more pressing. We’ve also had extensive evidence of AI wielding real-world harm long before that. The EU enacted comprehensive AI policy earlier this year, and we’re just a few months out from a heated election that has already seen AI being used to deceive and misinform voters. Considering all this, proposing AI regulation after a year of supposedly working to do so would hardly feel like rushing.
Just last week, I reported how the most comprehensive state AI bill yet bill blew up because of concerns that the federal government should be taking the lead (not to mention, opposition from the tech industry). At this rate, the states shouldn’t hold their breath.
And with that, here’s more AI news.
Sage Lazzaro
sage.lazzaro@consultant.fortune.com
sagelazzaro.com
AI IN THE NEWS
OpenAI sees two high-profile resignations—and Anthropic makes a big hire. Ilya Sutskever, OpenAI’s chief scientist and one of its cofounders, shared on X that he’s leaving the company after almost a decade to work on a “personally meaningful” project. Just hours later, Jan Leike, who was co-leading OpenAI's super alignment group along with Sutskever, shared a much shorter message: “I resigned.” The departures follow a series of high-profile exits from the company’s safety team over recent months, which has raised questions as to whether the team is gradually being hollowed out, as Fortune’s Christiaan Hetzner reported. Anthropic, however, has landed on the brighter side of this week’s AI executive shakeups. The company announced it hired Mike Krieger, who cofounded and served as CTO of Instagram and, more recently, AI company Artifact as its new chief product officer.
Stability AI is discussing a sale. That’s according to The Information, which reported the AI company has talked to at least one potential buyer in recent weeks. The four-year-old company was an early leader in generative AI with its Stable Diffusion model that produced realistic images from text prompts. The company has been grappling with the resignation of its founder and CEO, battles with investors, a staff exodus, and a serious cash crunch. Stability AI lost more than $30 million in Q1 and owes around $100 million to cloud providers.
Pulitzer winners tap custom models for investigative reporting as AI’s threats to publishers mount. City Bureau and Invisible Institute, which took home the award for local reporting for their “Missing in Chicago” project, said they trained a custom machine learning model to comb through thousands of police misconduct files. The New York Times visual investigations desk similarly trained a model to identify 2,000-pound bomb craters in Gaza throughout areas marked as safe for civilians. “We didn’t use AI to replace what would’ve otherwise been done manually. We used AI precisely because it was the type of task that would’ve taken so long to do manually that [it would distract from] other investigative work,” Ishaan Jhaveri, a reporter on the Times team, which won the award for international reporting, told Nieman Lab.
While the sheer number crunching and pattern recognition powers of AI are showing to be useful for journalism, other aspects of the generative AI revolution are starting to pose a major threat. Google this week announced it’s rolling out AI Overviews, a rebrand of its Search Generative Experience (SGE), in the U.S., and soon, globally. The product answers users’ Google searches with multi-paragraph answers, eliminating the need to click on website links for information in many cases. The test of the feature has already sent publishers scrambling and preparing for traffic declines ranging from 20% to 60% and up to $2 billion in revenue losses, Adweek reported.
FORTUNE ON AI
Google I/O shows why the search giant is sitting pretty against OpenAI —Sharon Goldman
Many CIOs are stuck in the AI slow lane but setting ambitious priorities as if they’re not —John Kell
Gen AI looks easy. That’s what makes it so hard —Rodney Zemmel (commentary)
AI CALENDAR
May 21-22: AI Seoul Summit on AI Safety in Seoul, South Korea
May 21-23: Microsoft Build in Seattle
June 5: FedScoop’s FedTalks 2024 in Washington, D.C.
June 25-27: 2024 IEEE Conference on Artificial Intelligence in Singapore
July 15-17: Fortune Brainstorm Tech in Park City, Utah (register here)
July 30-31: Fortune Brainstorm AI Singapore (register here)
Aug. 12-14: Ai4 2024 in Las Vegas
EYE ON AI NUMBERS
39%
That’s the percentage of CEOs who said they have good generative AI governance in place today. The number is not exactly encouraging, especially when you consider that 75% of the CEOs surveyed said trusted AI is impossible without effective AI governance in their organization.
There’s also the fact that these CEOs are rushing headfirst into bringing AI into their organizations. According to the survey, which was conducted by IBM and Oxford Business and consisted of interviews with more than 3,000 CEOs worldwide, 61% say they are pushing their organization to adopt generative AI more quickly than some people are comfortable with. And while 72% said they are currently no further than piloting and experimentation, 49% expect to be driving growth and expansion by 2026.
