David versus Goliath reimagine: OpenAI’s approach to AI supervision

Artificial general intelligence, or AGI, has people both intrigued and fearful.

As a leading researcher in the field, last July, OpenAI introduced the concept of superalignment via a team created to study scientific and technical breakthroughs to guide and ultimately control AI systems that are much more capable than humans. OpenAI refers to this level of AI as superintelligence.

Last week, this team unveiled the first results of an effort to supervise more powerful AI with less powerful models. While promising, the effort showed mixed results and brings to light several more questions about the future of AI and the ability of humans actually to control such advanced machine intelligence.

In this Breaking Analysis, we share the results of OpenAI’s superalignment research and what it means for the future of AI. We further probe ongoing questions about OpenAI’s unconventional structure, which we continue to believe is misaligned with its conflicting objectives of both protecting humanity and making money. We’ll also poke at a nuanced change in OpenAI’s characterization of its relationship with Microsoft. Finally, we’ll share some data that shows the magnitude of OpenAI’s lead in the market and propose some possible solutions to the structural problem faced by the industry.

Source