Another great interview. It is nice to drop into the middle of such a successful working partnership and friendship. I love the economic perspective on AI. Very fresh and interesting. I have been thinking about diffusion a lot these days. If you read any of the big AI briefs, you get the distinctive feeling that things are moving insanely fast, but perhaps that has more to do with the nature of contemporary media and our literal ability to keep up on everything (within limits of time and consciousness) through online access. That said, this same technical substructure seems to me to be the basis of a different kind of diffusion in this current AI moment. Cyber infrastructure seems like the X factor that is missing in the historical comparison these authors evoke. And yet we are trying to predict the future.. so really the conversation itself and how it influences present outlooks and policies is more of the essence.
Thanks Nick, glad you enjoyed it! Yes, I'm a big fan of Arjun and Zhengdong as people and as collaborators.
On diffusion and AI briefs, yes it does seem like things are moving fast, and it's also apparent if you spend time on AI Twitter / try to keep pace with the number of publications coming out any given day. I think, even in that picture, the number of publications / advances doesn't give a clear picture of how many of those will actually be useful, etc.
> That said, this same technical substructure seems to me to be the basis of a different kind of diffusion in this current AI moment. Cyber infrastructure seems like the X factor that is missing in the historical comparison these authors evoke.
By "different kind of diffusion," do you mean diffusion of knowledge? I'm also curious to hear you elaborate more on this point in general.
Yes, I was thinking diffusion of knowledge, but saying that I realize how anecdotal our accounts of speed of diffusion can be. Do technologies that allow for quicker diffusion of information lead to rapid increases in other technological development? There seem to me to be many intermediary causalities in between these two levels of diffusion that complicate the question.
I'd have to look for historical examples, but yes—you're right that there would be many intermediary causalities. It seems intuitive that quicker diffusion of information would lower some barriers to speedier technological development, though (speaking of diffusion lags, etc.) I would think there are other limiting factors (maybe access to internet, for instance) and that diffusion of information past a certain point would reach a saturation point in its ability to enable things like this.
Another great interview. It is nice to drop into the middle of such a successful working partnership and friendship. I love the economic perspective on AI. Very fresh and interesting. I have been thinking about diffusion a lot these days. If you read any of the big AI briefs, you get the distinctive feeling that things are moving insanely fast, but perhaps that has more to do with the nature of contemporary media and our literal ability to keep up on everything (within limits of time and consciousness) through online access. That said, this same technical substructure seems to me to be the basis of a different kind of diffusion in this current AI moment. Cyber infrastructure seems like the X factor that is missing in the historical comparison these authors evoke. And yet we are trying to predict the future.. so really the conversation itself and how it influences present outlooks and policies is more of the essence.
Thanks Nick, glad you enjoyed it! Yes, I'm a big fan of Arjun and Zhengdong as people and as collaborators.
On diffusion and AI briefs, yes it does seem like things are moving fast, and it's also apparent if you spend time on AI Twitter / try to keep pace with the number of publications coming out any given day. I think, even in that picture, the number of publications / advances doesn't give a clear picture of how many of those will actually be useful, etc.
> That said, this same technical substructure seems to me to be the basis of a different kind of diffusion in this current AI moment. Cyber infrastructure seems like the X factor that is missing in the historical comparison these authors evoke.
By "different kind of diffusion," do you mean diffusion of knowledge? I'm also curious to hear you elaborate more on this point in general.
Yes, I was thinking diffusion of knowledge, but saying that I realize how anecdotal our accounts of speed of diffusion can be. Do technologies that allow for quicker diffusion of information lead to rapid increases in other technological development? There seem to me to be many intermediary causalities in between these two levels of diffusion that complicate the question.
I'd have to look for historical examples, but yes—you're right that there would be many intermediary causalities. It seems intuitive that quicker diffusion of information would lower some barriers to speedier technological development, though (speaking of diffusion lags, etc.) I would think there are other limiting factors (maybe access to internet, for instance) and that diffusion of information past a certain point would reach a saturation point in its ability to enable things like this.