Eric Schmidt argues against a ‘Manhattan Project for AGI’

In a policy document, the former Google CEO Eric Schmidt, a scale EU Director General, and AI Security Director, Dan Hendrycks, AI must continue to continue the Manhattan Project Stool to develop AI systems.

A paper called “Super Management Strategy“The United States claims that only an aggressive proposal for managing super-sized AI systems can require a severe revenge in the form of a cyberaktak, which can destroy international relations.

“[A] Manhattan Project [for AGI] The opponents are likely to be able to see a denic or omnik to a denic or omnik, rather than acting to prevent this. “The thing that starts as a push for Superweapon and global control risks, the opposite and threatening the strain to strengthen the strategy and thus to ensure strategy.”

The author of the American AI industry is a few months after the author, paper, after a few months of paper The US Congress Commission offered the “Manhattan Project Style” In the 1940s, make an effort to finance modeled AGI development after the American Atomic Bomb Program. US Energy Secretary Chris Wright recently said, “he saidThe beginning of the new manhattan project“While standing in front of the SuperComputer site with Openai co-founder Greg Brockman in the EU.

Superting Strategy Paper, which has been a champion by several American policies and industry leaders in recent months, acknowledged that a program supported by AGI is the best way to compete with China.

Schmidt, Wang and Hendrycks are in the idea of ​​something that US not different mutually destroyed. The same way that global powers did not want a nuclear weapon in the same way – a holiday and co-authors who refuse an enemy – should be careful about the progress of its co-authored to extremely strong AI systems.

Extreme can be sounded in accordance with Nuclear Weapons, world leaders already consider the AI’s best military preference. Already Pentagon says that AI helps to accelerate the military’s killing chain.

Schmidt et al. A concept called a confidential AI Nazor (MAIM) is a concept that governments can actively disable the AI ​​projects that governments can threaten AI projects.

Schmidt, Wang and Hendrycks suggests that the United States has attracted its attention to developing the “competition race to win super management” methods Rape other countries from creating a superode AI. Co-authors claim the government to “expand [its] AI projects that threaten AI projects “AI projects managed by other peoples, as well as the arsenal of cybertakers to eliminate access to advanced AI plugs and open source models of enmity.

Co-authors determine a dichotomy that played in the world of politics. The EI development has “doomers, believing that there is a lawyer for the results of the preview of catastrophic consequences and the AI ​​progress. On the other hand, there are “ostriches” that people have to accelerate the development of the AI ​​and essentially not only to work.

Paper offers a third way: a measured approach to develop AGI, prioritizing defense strategies.

This strategy is noteworthy from Schmidt from Schmidt from Schmidt from Schmidt to the United States, which has previously developed aggressively compete with China in the development of previously developed AI systems. A few months ago, Schmidt left the word an op-ed Deepseleek, China and AI race marked a turning point.

Trump management seems to be dead against America’s development in the EU development. However, as co-authors noted, the decisions around the AGI of America do not have a vacuum.

The world’s watches show that the protection approach can be smarter, as America is pushing the threshold of AI, Schmidt and co-authors.