October 9th message, OpenAI CEO Sam Altmann, in a recent interview with Politico podcast Decoded, stated that the world should accept some of the negative events brought about by AI in exchange for technological gains and user autonomy。
One of the differences between OpenAI and some of the more stringent AI security advocates, he said, was that society, when exploring systemic resilience, "some bad things will happen" and that regulation should take a lighter approach。
This statement was made when discussing the difference between OpenAI and the AI security route of Anthropic。
Ottman argues that the benefits of AI will outweigh the disadvantages at the quantitative level, while Gary Marcus, Professor Emeritus at New York University, criticizes this statement for turning social risk into the cost of corporate expansion. Both positions revolve around the question of which risks can be assumed by society and which needs can be blocked before they are published, once models are more broadly open。
Otmann ' s focus is on placing regulatory boundaries on system resilience and user selection: if the risk can be detected, explained and repaired in use, he prefers to let the product enter the real environment first; if the risk causes irreversible harm, it should be limited before publication。
Marcus's rebuttal focused on the attribution of responsibility, arguing that “some bad things will happen” as a cost of innovation, and that the affected persons may be left to bear the losses that should be borne by companies and regulators。