Just last month, OpenAI disbanded its superalignment team. This internal unit, established in 2023, would help address long-term existential risks posed by artificial intelligence. If successful, this decision, which the CAG announces publicly in early 2024, will have a profound impact on the organization’s future approach to AI safety and alignment.
The superalignment team had a small group of six or seven members, who aimed to ensure that AI systems were “safe, trustworthy, and consistently aligned with human values.” Their cross-cutting work helped build foundational methodologies. These approaches are instrumental to AI following human intent robustly, especially in complicated real-world tasks and high-stakes settings. Allied by our passion for innovation and justice, our team came together in September 2023. Their purpose was to examine the best methods of managing and auditing AI.
Josh Achiam, who led the Alignment team, has moved into a new role as OpenAI’s Chief Futurist. He’ll be focusing on how the world adjusts to changes in AI and machine learning. He’s going to devote particular attention though to the effects of artificial general intelligence (AGI). Achiam has previously stated his desire to make sure that the long-term trajectory of humanity stays positive even with these rapid technological advancements.
“My goal is to support OpenAI’s mission — to ensure that artificial general intelligence benefits all of humanity — by studying how the world will change in response to AI, AGI, and beyond.” – Josh Achiam
As of this writing, Achiam’s personal web site includes him as being the lead of Mission Alignment at OpenAI. He quickly walked back his new position in a blog post. He will work closely with Jason Pruet, a physicist from OpenAI’s technical staff, to navigate the implications of AI advancements.
After the superalignment team was disbanded, the other members were spread out to different teams inside OpenAI. This definite change raises big questions about Achiam’s next steps. Will he build a new department or team in his futurist position? It’s anyone’s guess how he plans to address alignment concerns in the future.
According to Achiam’s LinkedIn profile, mission alignment director was a role he held from September 2024 until the recent organizational restructuring. OpenAI is taking another step in the right direction on the whole. It should remain fully committed to developing AI that is inherently controllable and ultimately aligned with human intent.
Platformer was the first to report the news about the disbanding of this superalignment team. This move away from military contracts further highlights OpenAI’s shifting approach toward AI alignment and safety.