Recently, Gary Marcus, a professor at New York University and a renowned AI scholar, publicly called for the market to immediately remove all connected open AI agents until their technical flaws and security risks are completely resolved.
Markus’s call stems from a serious safety incident caused by the Anthropic AI recently revealed by The New York Times. Considering various incidents of AI systems going rogue in the industry before, he points out that this generation of synthetic AI cannot be fully trusted by the public or businesses. In his view, these systems with security vulnerabilities should be recalled like cars with malfunctioning brakes and stopped from being sold in the market.
To further confirm the shortcomings in security control of large companies, Marcus also cited information provided by David Robinson, a former employee of OpenAI, during a recent media interview. In the interview, Robinson pointed out that the investment and control standards for safe operation environments among major cutting-edge AI companies, including OpenAI, remain at the “startup” level, and far from meeting the redundant standards required by nuclear power plants as expected by society and the public. He warned that the new technologies being developed today are not only more capable, but also pose an unprecedented risk of loss of control. If problems occur, the potential harm could far exceed that of a single nuclear power plant meltdown.
In addition, Marcus criticized the measures taken by relevant government agencies (such as the White House administration), which only require increased information disclosure, as being far from sufficient. This is tantamount to gently requiring criminals to declare their criminal records on their own every month. He emphasized that if the government fails to adopt stronger regulatory measures such as mandatory temporary recalls, and an uncontrollable major disaster occurs, both technology companies and the White House will be responsible for it.