Foundational models are like Public Goods (in economics): they add great value to the economy (like a lighthouse, a road), they cost a lot, but they are generally terrible private investment. So I want to give a big shout out to all the investors in these firms for helping out society in this way. But they'll lose their shirts. That does not in any way mean that applications of LLMs won't be incredibly powerful and great investments.

By analogy, this is like choosing to invest in building roads vs building automobiles. The US auto industry, and countless others did well from having a great highway system. We are thankful to President Eisenhower and others for helping to build that highway system.

Now, of course, we know the investors in LLMs are not mostly philanthropic in their objectives. They think they'll get the last laugh by inventing God (aka AGI) and instantly become owners of the rest of the world. But I think they are just naive in that regard. Even AGI will be competitive, just like we don't have just one genius human who rules all the others who are a little less capable.