OpenAI has announced the formation of a new Collective Alignment team, which will be responsible for integrating public input into the behaviors of its AI models. The team, comprised of researchers and engineers, aims to develop a system for gathering and encoding public feedback on OpenAI’s products and services.
Key Takeaway
OpenAI is taking steps to involve the public in shaping the behavior of its AI models through the formation of the Collective Alignment team, as part of its ongoing efforts to address regulatory concerns and ensure alignment with societal values.
Collective Alignment Team and Public Input
The Collective Alignment team is an extension of OpenAI’s public program, which was initiated in May to provide grants for experiments focused on establishing a “democratic process” for determining the rules that AI systems should adhere to. The company intends to collaborate with external advisors and grant teams, including conducting trials to integrate prototypes into steering their models. OpenAI is actively recruiting research engineers from diverse technical backgrounds to contribute to this initiative.
Public Program Recap
In a recent blog post, OpenAI highlighted the efforts of the grant recipients, which encompassed a wide range of projects such as video chat interfaces, platforms for crowdsourced audits of AI models, and methods to map beliefs to dimensions that can be utilized to fine-tune model behavior. The code utilized in the grantees’ work has been made public, along with brief summaries of each proposal and high-level takeaways.
Regulatory Scrutiny and OpenAI’s Response
OpenAI has faced increased regulatory scrutiny, including a probe in the U.K. regarding its association with close partner and investor Microsoft. The company has also sought to mitigate regulatory risks in the EU concerning data privacy. To address concerns about the potential misuse of its technology, OpenAI has announced its collaboration with organizations to limit the influence of its tools on elections through malicious means. This includes efforts to enhance the identification of AI-generated content and make it more apparent when images have been modified.
