Institute for Experiential AI: Towards a Foundation for AGI with Distinguished Lecturer, Gary Marcus

Summary

Best-selling author, entrepreneur and scientist Gary Marcus joined us as a Distinguished Lecturer host for 'Toward a Foundation for AGI' on September 28, 2022. Part of our Fall Seminar Series, Gary discsused many aspects of AGI: Large pretrained language models like GPT-3 and PaLM have generated enormous enthusiasm, and are capable of producing remarkably fluent language. But they have also been criticized on many grounds, and described as “stochastic parrots.” Are they adequate as a basis for general intelligence, and if not, what would a better foundation for general intelligence look like?

Notes

I really dig this talk. He talks a lot about the shortcomings of things like Chat GPT, arguing that if it cant figure out a blue block from a red block or do simple math then there is no way that we will reach AGI by 2029 (as Elon Musk predicted). He even offered Elon a $500,000 bet that it wasn't gonna happen.

Instead, he offers the notion that we need to include reasoning and human values in order to reach the level of Artificial Intelligence that is necessary for AGI. In direct opposition is the fact that it has no idea what's going on in the world, and hasn't a clue what life or death mean; he even cited an example that included a bot responding, telling someone to commit suicide could be a major concern, what's to stop it if it concludes that it would likely result in the greatest level of happiness.

• He argues that we're at worst time in AI history because there are so many risks now that we are using it.
• This guy is getting into arguments with the listeners, what's the controversy #❓
• Really drives home the idea of Neuro Symbolic.
• Claims robotics aren't very impressive and unnecessary, but would be helpful.

Action

• #❓ #on/ai #to/learn What is Gofi? or GOFAI? Is it Good Old Fashion AI?
• #❓ #on/ai #to/learn Look into Neuro Symbolic (learning?)

up:: 📥 Sources