Senator Bernie Sanders and Representative Greg Casar officially announced the Ban Artificial Superintelligence Act. The legislation is aimed at significantly slowing down the pace of advanced AI developments through federal prohibitions. Its fundamental objective is to protect the public from existential, structural and social risks of advanced computing. The legislation also creates the Department of Artificial Intelligence, a new, Cabinet-level federal agency led by a secretary appointed by the President. The Department of AI will be composed of independent scientists and technical experts.
What is the “Ban Artificial Superintelligence Act”?
The bill outlaws artificial superintelligence. It defines superintelligence as any AI system that matches or exceeds human cognitive performance across a range of tasks. It also includes systems capable of undermining the US government or enabling the disempowerment of humanity. This definition expands to any system that possesses the capacity to overthrow human control, evade shutdown commands and execute unauthorized cyberattacks.
It also calls for immediate cessation of frontier AI models currently being developed. This pause is to remain in effect until the Department of AI is able to establish a rigid framework for federal approval and safety reviews. This will stop companies from deploying new advanced models without state approval. The Department is also tasked with tracking frontier AI models from initial training to deployment. If the Department concludes that the model is developing dangerous capabilities, it has the authority to force the removal of such features and even supervise the complete destruction of the code.
The bill also calls for a “corporate death penalty”. Any tech executives who do not adhere to these rules are to face up to 20 years in prison. Moreover, corporations that attempt to circumvent safety review processes or build banned systems are at risk of losing their legal charters and ability to operate.
President Trump’s encouragement of superintelligence one day before the bill
Just a day before this bill was unveiled, President Trump while addressing the UN General Assembly, announced that he wanted to officially rebrand “artificial intelligence” to “super intelligence”. His basis for this change was that “artificial” suggests something is fake, and undermines AI’s credibility. He further emphasized that he would encourage superintelligence and protect it from regulations. The timing of this bill creates an interesting contrast. Bernie Sanders is aiming to ban the exact term that President Trump used to promote the technology.
President Trump’s resistance to the bill is likely to be viewed as a necessary defense of American innovation and national security as a means of competing with adversary states. His approach will be backed by Silicon Valley allies as a pragmatic approach. Critics of the bill argue that pausing American AI development equates to surrendering before China.
In a Republican-controlled Congress, this bill is likely to receive no support. The creation of the Department of AI with the mandate to liquidate companies is viewed as detrimental to American progress.
