The team is a result of programme launched last May that awarded 10 grants worth €100,000 each to fund experiments into ‘democratic inputs to AI’.
OpenAI has shared plans to put together a team of researchers and engineers that will crowdsource public opinion on the behaviour of its AI models and implement suggestions.
According to a statement yesterday (16 January), the Collective Alignment team aims to design systems that incorporate public inputs to steer models while addressing challenges ranging from a digital divide and polarised groups to the representation of diversity and “anxieties” around the future of AI governance.
The AI start-up said the new team will work with external advisors and grant teams, including running pilots to incorporate the grant prototypes into steering its models.
“We are recruiting exceptional research engineers from diverse technical backgrounds to help build this work with us,” OpenAI wrote, sharing a link for interested individuals to apply.
The Collective Alignment team is a result of OpenAI’s public programme launched last May that awarded 10 grants worth €100,000 each to fund experiments around the world into “democratic inputs to AI”.
The company said at the time that the goal of the programme was to create a democratic process for deciding what rules AI systems should follow.
“By ‘democratic process’, we mean a process in which a broadly representative group of people exchange opinions, engage in deliberative discussions and ultimately decide on an outcome via a transparent decision-making process,” OpenAI wrote in a blog post in May.
“Some participants felt nervous about the use of AI in writing policy and would like transparency regarding when and how AI is applied in democratic processes,” OpenAI wrote. “Post-deliberation sessions, many teams found that participants became more hopeful about the public’s ability to help guide AI.”
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.