China's cheaper and open-weight models leave the closed and more expensive U.S. models vulnerable to competition
China may now be leading the artificial-intelligence race from behind.
As the former chief executive officer of Google, Eric Schmidt is well-positioned to assess the state of play between the U.S. and China in the battle for AI superpower status. The good news? Right now, he reckons China is about six months behind. The bad news? If that's true, China is already winning because its models are either cheaper or free, and the vast majority of the world will adopt those.
This troubling narrative was outlined by Schmidt during a podcast interview released Wednesday by the chief executive of Norway's $2 trillion-dollar sovereign wealth fund, Nicolai Tangen. Schmidt, who led Google between 2001 and 2011, is now the chair and chief executive of an aerospace manufacturer called Relativity Space while also investing in tech personally. He still holds nearly 16% of Google parent Alphabet (GOOGL) (GOOG), according to FactSet, and Bloomberg has reported that he owns 1% of the AI lab Anthropic.
Eric Schmidt.
Schmidt's concern is that, because China's models are open weight, they will clearly be more appealing to consumers in less-wealthy countries who cannot afford the more expensive, closed Western models.
His hunch is borne out by data. The AI routing company OpenRouter says the five most popular models are Chinese.
Schmidt points out that the rapid advances made by the Chinese probably have been made because "they have used chips that they were not supposed to have" but were "pilfered or smuggled" into China. At the same time, he says, China has exploited the distillation technique - posing questions to a Western model and then training their own models on the answers.
Worryingly for European governments, though, Schmidt is scathing on their prospects and how far behind the U.S. and China they have fallen. Schmidt cites a risk-averse culture; high energy costs; low levels of university funding; and, above all, overregulation. He tells Tangen that Europeans are change-resistant, and that "they regulate before they innovate."
Comparing the approach of Europeans to Americans in the evolution of AI, Schmidt notes that, when Brussels designed its strategy five years ago, it was prefaced by a list of things that could not be done. The American tactic was "run as fast as you can, align it with human values, and, if bad things happen, fix them immediately."
As one might expect from someone who owns a personal stake in Anthropic (worth maybe $20 billion), Schmidt is an AI bull. During the podcast conversation, he makes an immense effort to convince Tangen of the size, scope and speed of the AI revolution, which he thinks is beyond the capacity of most people to fathom. "It defies logic," he explains. "It's like the scale of the universe - it's too big for humans to understand."
On the boom-or-bubble question, Schmidt is emphatic: "This is not a bubble." He cites the revenue growth of Anthropic as evidence: "When I was CEO at Google, we set the revenue-growth record. [Anthropic] smashed it [with] more than twice the speed we [had]. That's real growth. It's not fake. It's real money, with real customers who are paying for something useful."
Citing a draft prospectus, the Wall Street Journal reported Anthropic generated $16.3 billion of revenue in the first half of the year, alongside a net loss of more than $80 billion.
Schmidt waxes evangelical on the benefits AI may bring to the study of biology, mathematics and physics. He quotes the current chairman and co-founder of Google DeepMind, Demis Hassabis, who has publicly predicted that in the next few decades all human disease can be cured.
On a company level, Schmidt, who led Novell before making the jump to Google, thinks chief executives who adopt and harness AI fastest will be the biggest winners. The emphasis is on the speed of the take-up because the industry is moving so fast, development cycles are being compressed all the time and some models are obsolete within six months.
His advice to chief executives is to flatten organizations, eliminate "glue people" who perform routine tasks, and focus on inventors, salespeople and leaders.
-Jules Rimmer
Comments