Superalignment is research aimed at steering potentially superintelligent AI toward human intentions and values, preventing systems from going rogue. Its urgency follows from the possibility that superintelligence could emerge this decade. section 1.1
A four-year safety challenge
OpenAI launched a Superalignment team co-led by Ilya Sutskever and Jan Leike. The team set itself four years to address how a potentially superintelligent system could be steered safely, turning alignment from a general aspiration into an organizational research priority. section 1.1
