Using graphs for AGI
AA: I’d like to get your thoughts on AGI. Do you think graph representation learning is one of the paths forward? What about algorithmic reasoning, another area that you’re working on? How do you see the field of AGI emerging? Do you think it’s something that’s attainable?
PV: That’s a very important question, and one that we must be increasingly mindful of as we become more powerful in building these kinds of systems.
Should we be using graph representation learning as part of AGI? I wholeheartedly think we should be. Think about a truly generally intelligent system: what does such a system (likely) do? It first extracts meaningful concepts out of all of that richness of real-world signals. How does such a model then know how to make conclusions and go forward in their life, so to speak? They do so by taking all the concepts, relating them to past experiences, and using reasoning arguments to derive new pieces of knowledge...