Noam Shazeer: belief

Transcript context

…I think we're an "organize the world's information" company, and that's broader than information retrieval. Maybe: “organizing and creating new information from some guidance you give it”. "Can you help me write a letter to my veterinarian about my dog? It's got these symptoms," and it'll draft that. Or, "Can you feed in this video, and can you produce a summary of what's happening in the video every few minutes?" I think our multimodal capabilities are showing that it's more than just text. It's about understanding the world in all the different modalities that information exists in, both human ones but also non-human-oriented ones, like weird lidar sensors on autonomous vehicles, or genomic information, or health information. And then, how do you extract and transform that into useful insights for people and make use of that in helping them do all kinds of things they want to do? Sometimes it's, "I want to be entertained by chatting with a chatbot." Sometimes it's, "I want answers to this really complicated question, there is no single source to retrieve from." You need to pull information from 100 web pages, figure out what's going on, and make an organized, synthesized version of that data. Then dealing with multimodal things or coding-related problems. I think it's super exciting what these models are capable of, and they're improving fast, so I'm excited to see where we go. I am also excited to see where we go. I think definitely organizing information is clearly a trillion-dollar opportunity, but a trillion dollars is not cool anymore. What's cool is a quadrillion dollars. Obviously the idea is not to just pile up some giant pile of money, but it's to create value in the world, and so much more value can be created when these systems can actually go and do something for you, write your code, or figure out problems that you wouldn't have been able to figure out yourself. To do that at scale, we're going to have to be very, very flexible and dynamic as we improve the capabilities of these models. Yeah, I'm pretty excited about a lot of fundamental research questions that come about because you see something could be substantially improved if we tried this approach or things in this rough direction. Maybe that'll work, maybe it won't. But I also think there's value in seeing what we could achieve for end-users and then how can we work backwards from that to actually build systems that are able to do that. As one example: organizing information, that should mean any information in the world should be usable by anyone, regardless of what language they speak. And that I think we've done some amount of, but it's not nearly the full vision of, "No matter what language you speak, out of thousands of languages, we can make any piece of content available to you and make it usable by you. Any video could be watched in any language." I think that would be pretty awesome. We're not quite there yet, but that's definitely things I see on the horizon that should be possible.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.