OpenAI, the American AI company, has announced that it is offering $10 million in grants to researchers who can help ensure the safety of superintelligent AI systems. The company aims to understand how current AI models can be used to monitor and assess the outputs of more advanced AI systems.
Through its Superalignment Fast Grants program, OpenAI hopes to gain insights into controlling AI systems that surpass human intelligence. Additionally, the company aims to explore the possibility of building an AI lie detector.
OpenAI acknowledges that fully understanding superhuman AI systems will be a challenging task. Humans cannot reliably evaluate whether millions of lines of complex AI-generated code are safe to run. Therefore, the company seeks to incentivize research in this domain by offering grants to individual researchers, non-profits, and academic labs.
The company is also introducing the OpenAI-sponsored $150,000 Superalignment Fellowship for graduate students interested in alignment research. Prior experience in alignment is not a requirement for the fellowship, and applications are open until February 18.
OpenAI emphasizes that ensuring the safety of future superhuman AI systems is a crucial and unsolved technical problem. The company believes that new researchers can make significant contributions to addressing this challenge.
To keep AI systems safe and accountable, OpenAI has identified seven practices, and it now plans to fund research that addresses open questions arising from previous studies. Part of this effort includes awarding grants ranging from $10,000 to $100,000 for research on the impacts of agentic AI systems and practices for ensuring their safety.
Agentic AI refers to superintelligent AI systems capable of autonomously performing various tasks and acting on complex goals on behalf of users. OpenAI believes it is essential to make agentic AI systems safe by minimizing failures, vulnerabilities, and potential abuses.
This commitment to safety comes as researchers have recently discovered that cybercriminals can exploit OpenAI’s AI-powered chatbot, ChatGPT, for malicious purposes.
OpenAI’s dedication to advancing AI safety research underscores its mission to ensure that humans can retain control and proper oversight of future superintelligent AI systems. By supporting groundbreaking research, the company aims to make significant progress toward aligning AI systems with human values and creating a safe and beneficial AI-powered future.
