(Bloomberg) -- While President Biden was in San Francisco finalizing an agreement between the US and China to open a dialogue on artificial intelligence, among other issues, a who’s who of AI and military leaders gathered nearly 800 miles away for a secretive event in the Utah mountains.Over the course of three days last week, more than 100 AI executives and insiders, venture capitalists and government officials listened to talks about the AI arms race between the US and China, and mulled the impact of Biden’s sweeping executive order on the industry. They discussed how AI could be used by bad actors and considered the next possible tectonic shifts in the industry. They also indulged in s’mores and ventured into the snow-dusted terrain at the Montage Deer Valley resort in Park City to learn to shoot with a bow and arrow.The event, called the AI Security Summit, was hosted by Scale AI Inc., a startup that offers AI training and data labeling services to customers ranging from OpenAI and General Motors Co. to the US Army. Only those invited had reason to know about it. An all-black, password-protected webpage for the summit gave little away. “Join the world’s brightest AI leaders and visionaries,” a message on the site read. “By invitation only.”For the AI community, the summit served as a chance to talk through uncertainties ahead — around AI regulation, political conflicts and the technology itself — in a luxury retreat far removed from hearing rooms in Washington and boardrooms in Silicon Valley. It also offered an opportunity to speak more freely than at public events, as attendees agreed to only share information from the conference without referencing anyone’s identities or place of work. For Scale, the summit could help burnish its reputation as a leader of such discussions and position it at the nexus between the AI and defense worlds. Scale rented out a hotel for the three-day event and footed the bill for most attendees. Bloomberg attended as a paying guest.Scale agreed Bloomberg could name some speakers who attended the event. They included Matt Knight, head of security at OpenAI; Andrej Karpathy, an OpenAI researcher and Tesla’s former director of AI; Craig Martell, the Pentagon's chief digital and artificial intelligence officer; and General James Rainey, the commanding officer of the US Army Futures Command. Also in attendance was Alexandr Wang, Scale’s chief executive officer and co-founder, who has been called “Washington’s AI whisperer,” and has been particularly vocal on highlighting threats posed by China.
Wang said the focus of the conference was informed by rising geopolitical tensions — never mind that the event took place right as leaders from the US and China were meeting to ease tensions. It was also spurred by the dramatic increase in capabilities and adoption of AI in the year since OpenAI’s ChatGPT launched. “This past year has been absolutely staggering for anyone in the field,” he told Bloomberg News. “More has happened in the past year than happened in the prior 10.”For all the focus on looking at future opportunities and risks for the industry, there was no sense of a more immediate change coming to the best-known AI startup. Days after the conference, Sam Altman, the chief executive officer of OpenAI and face of the artificial intelligence industry, was ousted from his position. There were no public indications from Karpathy or Knight at the time that anything was amiss at the company.One panel discussion focused on AI competition between the US and China, with signs of concern over China’s posture toward Taiwan, the world’s semiconductor capital. Another talk offered a deep dive into the intricacies of the chips and hardware needed to power AI, with a panelist breaking down the difficulties of building data centers fast enough to satisfy customers’ rapidly increasing demands. A third discussed how companies can secure their most powerful AI models. There was quite a bit of optimism throughout. Over avocado toast at breakfast and berry smash cocktails before dinner, attendees marveled at the pace of AI innovation. They also questioned each other on the likelihood of AI advancing to a point where it outperforms humans and discussed whether the technology presents an existential risk. While these discussions can prompt vitriol on social media, one attendee noted the in-person chats were more nuanced and reasonable.
Raquel Urtasun, CEO of autonomous driving startup Waabi and a speaker at the event, said she chatted with others about how large AI models will continue to scale and how AI will increasingly impact the physical world, including with the rise of self-driving vehicles.When not in panels, the attendees enjoyed a little tech-industry excess, executed with local style. They relaxed at a sprawling compound billed as a “refined mountain retreat,” picked out custom cowboy hats and coordinating hat bands, and dined on coffee-rubbed bison tenderloin served with wild mushrooms and blueberry jus.
During an archery session on the second day, guests divided into small groups and attempted to hit bullseyes and a target that looked like a furious, life-sized bear, as well as some strategically-placed balloons. After observing a teammate’s good shot, one guest made a distinctly AI-themed joke to his group: “He didn’t need a lot of training data.”
©2023 Bloomberg L.P.