champaignmagazine.com

By W.H.L.

OpenAI’s release of GPT-4o on May 13, 2024 is another milestone towards AGI. Recently Ray Kurzweil reiterated his prediction that AI will reach the human level of intelligence in 2029, and Musk said in reaction that it will probably happen next year in 2025.

While it would be nice if a prediction turns out to be accurate, it doesn’t necessarily help the matter much. If AGI is inevitable and near, how should human creators walk the last mile and get there with safe and responsible AGI that will benefit the whole humankind? How should AGI user community and society be prepared for that AGI-powered reality?

Those are open-ended questions: we may not have answers right now, but we need to think about those urgent “how” questions.

How to get there? This could be more a strategic question than a technical one. The exodus of some key members of former OpenAI Superalignment team confirms the fact that there are different opinions and concerns about the goals, roadmap, and priorities to match AGI development with human interests. It would help the AI community and general public to better understand the importance of AGI’s reliability and possible vulnerability, if OpenAI and the critics can openly discuss their concerns from each side’s perspectives.

How should humans treat AGI as smart as us? I don’t think we know how. We are currently not well prepared to confront that as reality. In 2012 David Deutsch wrote in The Guardian:

“I am convinced that the whole problem of developing AGIs is a matter of philosophy, not computer science or neurophysiology, and that the philosophical progress that will be essential to their future integration is also a prerequisite for developing them in the first place.”

More than a decade later, in 2024, little progress has been made to fulfill that prerequisite and lay a solid foundation for AGI. Without that philosophical framework to guide us, how do you expect AGI and humans to co-exist rationally and peacefully?