Former Google CEO Eric Schmidt warned that the aggressive pursuit of AI development and weaponization could pose new risks. Here’s what he meant.
Former Google CEO Eric Schmidt warned that the aggressive pursuit of AI development and weaponization could pose new risks. Here’s what he meant.
«What begins as a push for superweapons and global control risks provoking hostile countermeasures and escalating tensions, thereby undermining the very stability the strategy is supposed to provide,» Eric Schmidt notes in an analytical paper titled «The Strategy for Superintelligence,» which he wrote in collaboration with Scale AI CEO Alexander Wang and Director of the Center for Artificial Intelligence Security Dan Hendricks, Forbes reports .
The document opposes the so-called «Manhattan Project» approach to developing artificial general intelligence (AGI). Parallels are drawn with the Manhattan Project, a research and development program launched during World War II to create the first nuclear weapons. At that time, it was led by the United States in cooperation with Great Britain and Canada. Recently, a US congressional commission proposed an attempt to fund the development of AGI in the style of the «Manhattan Project».
The authors of the document warn that the aggressive US-led pursuit of superintelligent artificial intelligence systems could provoke harsh retaliation from China, potentially destabilizing international relations.
«[The Manhattan Project] [for AGI] assumes that adversaries will tacitly acquiesce to sustained imbalance or genocide rather than act to prevent it,» the co-authors write in «Strategy for Superintelligence.»
Eric Schmidt and his associates warn the United States and advise it to exercise caution when trying to dominate powerful AI systems.
Schmidt, Wang, and Hendricks propose shifting the focus from «winning the superintelligence race» to developing methods that deter other countries from creating superintelligent AI. They introduce a concept called Mutual Assured AI Malfunction (MAIM), in which governments can shut down threatening AI projects early, rather than waiting for enemies to weaponize AGI.
