As seen by Wang’s attendance on Sunday during the first night of Web Summit Qatar, the statement generated a range of responses. Felix Salmon of Axios, Wang’s interviewer, asked the audience how many people shared that viewpoint, and he only counted two hands. Salmon reported that an “overwhelming” number of hands went up when he asked the group how many people disagreed.
So Salmon asked Wang and explained. “AI is going to fundamentally change the nature of national security,”. He stated that he grew up in Los Alamos, New Mexico “the birthplace of the atomic bomb”. Both of his parents were physicists and worked in the National Lab.
Wang said, reflecting his own thoughts on the U.S.-China competition. The full-page ad was sparked by his concern that AI might enable China to “leapfrog” the military might of “Western powers.”
Wang was repeating words that are increasingly coming from venture capitalists and defense tech businesses. They are advocating for more AI weapons in general as well as increased autonomy in AI weapons. They cite China and imagine a scenario in which China develops completely autonomous AI weapons while the United States is delayed by the need for a human decision-maker to be involved before shooting.
Wang attempted to argue for the choice between China and the U.S. for baseline LLM models, going beyond the hypothetical weapons of another country.
Indeed, scholars have found that the censorship of the Chinese government is ingrained in a number of well-known LLM models. The Chinese models are also beset by worries about data collection backdoors used by the Chinese government.
Given that Scale had just announced its agreement with the Qatari government, Wang’s fears of political involvement in AI were particularly pertinent. The scale will assist Qatar in developing 50 AI-powered government applications, spanning from healthcare to education, Wang said in an announcement made on Sunday.
Scale is famous for deploying large numbers of contract workers, often from overseas, to help train models by hand. In addition to Microsoft, OpenAI, and Meta, it works with most of the major U.S. foundational models.
Scale AI’s overtly pro-American rhetoric probably appeals to its DoD clients. However, the Web Summit discussion also demonstrated how many people appear to be equally uneasy about the United States possessing superpowers in AI.
