Korva is a scroll of the internet’s best ideas.
Too dangerous to distribute
The first camp, Anthropic and the effective altruism movement, treats AI as comparable to a nuclear weapon. Its vision keeps chips and models inside the United States, sets up something like an atomic energy commission to regulate the industry, and ties a few companies closely to the government. It also expects competition to create a race to the bottom on safety.
Too dangerous to centralize
The second camp, which includes Mark Zuckerberg, Elon Musk, and Jensen Huang, holds that history favors distribution. Hoping that an absolute power will benevolently provide for humanity has not led to safe outcomes, and there is no counterexample where centralizing worked better. It expects competition to bring out the best, and it wants individuals to be able to own an AI whose values are aligned with theirs.
A consumer technology
No individual needs a nuclear weapon, but every consumer and every business needs AI, which makes it a consumer technology before it is a military one. The technology also spreads regardless, and banning open source or centralizing control would hand AI to China.
How did the US regulate nuclear power?
Congress passed the Atomic Energy Act in 1946, creating the Atomic Energy Commission to control nuclear materials and research. A 1954 revision allowed private industry to use nuclear technology under licenses. In 1974 the commission was split, with the Nuclear Regulatory Commission taking over safety licensing. Internationally, the International Atomic Energy Agency was founded in 1957.
What happened when governments tried to control encryption?
In the 1990s the US classified strong encryption as a munition under export rules. Phil Zimmermann released the PGP program in 1991, and its source code was later published as a printed book that could be exported legally. The government moved oversight of encryption exports to the Commerce Department in 1996 and relaxed the rules substantially in 2000. Strong encryption is now built into web browsers and phones.
How do checks and balances create safety?
James Madison argued in Federalist No. 51 in 1788 that a government of competing powers could restrain itself, because each branch's ambition would check the others. The US Constitution splits authority among Congress, the president, and the courts for that reason. The idea has older roots in Montesquieu's The Spirit of the Laws, published in 1748. Madison drew heavily on it while writing.
More from All-In with Chamath, Jason, Sacks & Friedberg
