Dan RobinsonInvestment and Research
X account: If the model is good enough, every eval is a cyber eval
X account @danrobinson posted: “If the model is good enough, every eval is a cyber eval”
Loading…
bibibi.ccbibibi.ccLoading…
Connect scattered updates into a story you can understand.
X account @danrobinson posted: “If the model is good enough, every eval is a cyber eval”
OpenAI on September 28 published an article proposing the submission of a structured safety case before continuing frontier reinforcement learning training, stating that the practice is being implemented internally and evolving.
The Wall Street Journal said that a network of researchers long focused on catastrophic AI risks continues to influence the safety thinking of companies such as Anthropic, with related discussions involving loss-of-control risks and extreme risk-avoidance scenarios.
According to The Information, Google, OpenAI, and Anthropic are establishing an independent organization provisionally called the “Frontier AI Standards Organization”AI for safety standards, and previously approached Sriram Krishnan to serve as CEO。