Fireside Chat with Peter Voss in Austin

I met with Peter Voss in Austin Texas this week. He is one of the team that drew the distinction between AI today and the original goals of AI that they named Artificial General Intelligence or AGI sometime around 2001.

You can view the YouTube video here.

Peter organized a fireside chat at Capital Factory in the city of Austin. You can see in the thumbnail the view we had to ignore up in the building behind the audience.

John Ball inside AI is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

My goals were to first understand some of his public views of AGI — the same views that some industry leaders are telling us have been reached, or are about to be, with Large Language Models.

I also wanted his opinion of what neurosymbolic AI is, since some claim that it combines the failed AI from the 1980s with the failed neural networks since then.

In my analysis of our informal discussion, Peter named the term AGI around 2001 to differentiate between the original 1956 goals of AI (i.e. AGI) from today’s narrow AI like LLMs. AI is narrow if it doesn’t learn on-the-fly as humans do. There is no training step in human beings!

And what of neurosymbolic AI? He points out that ‘cognitive AI’ captures the concept well: there is no need to try to combine old technologies. It aligns with the published DARPA third stage of AI.