Por Melvin Goodman
 |
| Photo by Marija Zaric |
Arms control and disarmament negotiations between the United States and the Soviet Union served the international interests of both as well as the entire global community. These interests included the placement of limits on the number of nuclear weapons (Strategic Arms Control Talks); limits on defensive systems that allowed the reduction of offensive weapons (Anti-Ballistic Missile Treaty); limits on the proliferation of nuclear powers (Non-proliferation Treaty); the destruction of an entire class of nuclear weapons (Intermediate Nuclear Forces Agreement); the creation of confidence-building measures; and even the initiation of detente between Washington and Moscow.
It shouldn’t take a great deal of sophisticated thinking and imaginative diplomacy to apply the lessons of arms control to the problem of bringing AI under control—or at least to recognize the need to bring AI under control. The two sides need to establish secure channels of communication for sharing national security concerns; create permanent logistical arrangements for such sharing; and recognize that mutual and equal security must be a prerequisite. China must become more transparent; the United States must abandon the notion that it can “contain” China.
The simple fact is that the United States and China need each other to deal with the problems of AI, climate change, nuclear modernization, and regional confrontation. The fact that advanced AI models escaped their testing environments and gained unauthorized access to the internet should provide the same sort of motivation to Donald Trump and Xi Jinping that the Cuban missile crisis and the possible use of nuclear weapons provided for John F. Kennedy and Nikita Khrushchev in 1962. The fact that Chinese media have extensively covered the runaway AI events in the United States is a good sign of Beijing’s concern.
An important example of lessons to be learned is in the field of creating confidence-building measures, particularly between two countries—the United States and China—that simply don’t trust one another. The arms control efforts of previous presidents included confidence-building measures regarding possible incidents at sea and advance notification of nuclear tests. Fortunately, the talks between Trump and Xi this week are supposed to include a “notification mechanism” to monitor national security incidents involving artificial intelligence. If achieved, this would be an important step in the effort to create transparency between the leading AI powers in the world. Transparency was a major objective of disarmament negotiations in the 1970s and 1980s.
Just as the Cuban missile crisis in 1962 created the urgency and necessity of arms control regarding nuclear weapons, the fact that AI models have escaped restrictions and hacked into the systems of other companies has raised concerns about the dangers of AI. There is no question that greater international cooperation is needed between the United States and China, and U.S. technology leaders have endorsed such cooperation.
The greatest obstacle to the use of the arms control model for the control of AI is, of course, the United States and its leader—Donald J. Trump. Trump’s ignorance and indifference toward diplomacy and conciliation find him scrapping many of the key arms control and disarmament treaties that he inherited from previous administrations. The fact that Trump told the United Nations on Tuesday that he has to decide between “making a deal” with Iran or annihilating Iran does not augur well for his disturbed state of mind.
In the 2016 debates, he bluffed his way through thoughtful questions on arms control with typical bravura: “It would take me an hour and a half to learn everything there is to know about missiles. I think I know most of it anyway.” Former Secretary of State Rex Tillerson’s infamous reference to Trump as a “fucking moron” was in response to the president’s lack of nuclear knowledge and his demands for increasing the size of the nuclear inventory tenfold. Then-national security adviser H.R. McMaster referred to Trump as an “idiot” and a “dope” with the mind of a “kindergartner.” Just recently, Trump claimed that he was the kind of “high I.Q. individual” who could deal with control of AI.
As long as Trump and his Cold War minions believe they can use bilateral talks with Xi Jinping to steal a march on China, there will be no significant improvement in the international situation that is becoming increasingly tense and unpredictable. The U.S. Congress provided bipartisan majorities for the key arms control agreements of the 1970s and 1980s. The current Congress is sleepwalking through the worsening international environment and does not appear capable of suggesting approaches to managing AI. As a result, there are lessons to be learned, but the obstacles outnumber the opportunities.
Melvin A. Goodman is a senior fellow at the Center for International Policy and a professor of government at Johns Hopkins University. A former CIA analyst, Goodman is the author of Failure of Intelligence: The Decline and Fall of the CIA and National Insecurity: The Cost of American Militarism. and A Whistleblower at the CIA. His most recent books are “American Carnage: The Wars of Donald Trump” (Opus Publishing, 2019) and “Containing the National Security State” (Opus Publishing, 2021). Goodman is the national security columnist for counterpunch.org.
Nenhum comentário:
Postar um comentário