OpenAI has announced plans to make substantial investments and establish a dedicated research team, called the Superalignment team, to prioritise the safety of its AI. The company aims to ensure that AI systems remain safe and beneficial for humans. Notably, OpenAI envisions utilising AI itself to oversee and supervise its own operations in the future.
The company aims to dedicate 20% of its computing power over the next four years to solve the challenge of controlling superintelligent AI. Their strategy involves training AI systems using human feedback, training them to assist with human evaluation, and ultimately training AI systems to conduct alignment research themselves. However, some AI safety advocates, like Connor Leahy, argue that the plan is flawed since initial human-level AI could cause havoc before being compelled to solve AI safety problems.
The potential dangers of AI have gained widespread attention, leading to calls for caution and responsible development. In April, industry leaders and experts called for a six-month pause in developing more powerful AI systems, citing societal risks.