Two hundred major IT businesses have joined a US-based initiative to promote ethical AI practices. As members of the US AI Safety Institute Consortium (AISIC), Meta, Google, Microsoft, and Apple will be included. Gina Raimondo, the secretary of commerce, recently revealed the group’s large number of new members and stated that they will be responsible for implementing the measures outlined in President Biden’s extensive executive order on artificial intelligence.
“In order to reduce the risks and fully utilize the enormous potential of artificial intelligence, the U.S. government must play a major role in establishing the guidelines and creating the necessary resources,” Raimondo stated in a statement.
The phrase “red-teaming” in cybersecurity has its roots in the Cold War. It alludes to computer games where the adversary was referred to as the “red team.” An AI that is determined to behave badly would be the adversary in this scenario. Those that engage in this behavior will attempt to use prompt hacking to fool the AI into doing malicious activities, such as disclosing credit card numbers. People can create more effective defenses once they understand how to breach the system.
Another crucial component of Biden’s initial directive is watermarking synthetic content. Members of the consortium will create policies and procedures to guarantee that users can recognize content produced by artificial intelligence. Hopefully, this will reduce the spread of false information boosted by AI and deep fake tricks. Although the program will “facilitate and help standardize” the underlying technological requirements underpinning the practice, digital watermarking has not yet gained widespread acceptance.
Though the Commerce Department claims the consortium constitutes the greatest collection of testing and evaluation teams in the world, the consortium’s work is only getting started. Right now, all we really have is this connected consortium and Biden’s executive order. Congress consistently fails to enact any form of significant AI legislation.