Three researchers who previously worked in the field of AI safety at OpenAI and were subsequently dismissed have warned that oversight should not be relaxed for increasingly autonomous AI systems.
According to a report by the Wall Street Journal, the researchers expressed their concerns in a letter sent to members of the OpenAI board of directors, stating that companies may eventually lose the capacity to monitor the reasoning processes of their AI systems.
The letter emphasized that security audits against the risks posed by autonomous AI systems should not be limited to internal company mechanisms but should instead be conducted by independent organizations.
The researchers stated that there is currently no clear method in the industry for how to safely develop and deploy models that cannot be audited. For this reason, they called on OpenAI and other leading AI companies not to advance technology in a way that makes oversight more difficult.
A spokesperson for OpenAI told the newspaper that they fully agree with the suggestions in the letter and that the auditing of AI models is of great importance to the company.
It is alleged that the researchers in question were dismissed on the grounds that they shared confidential information with an AI safety company.
News Source: 12punto
