Felix Oberholzer-Chi

One of the most remarkable things is I would ask different questions, like give me a list of M&A targets, like those kinds of things.

And very often the answers are spot on or like, do you think there is excess supply of office space in the United States?

And the answer you get is just unreal that a machine can answer that.

But then what I also found is if the answer is profoundly wrong, it also sounds very convincing.

Yeah, super confidence.

It's almost like a person with no filter.

It can be completely misleading.

But I wonder, in these dinner party conversations, I often find when you then push a little bit

You will eventually, I think, get to a set of assumptions that the person makes, a particular set of experiences that are maybe very different from your experiences.

And you come to see the world very differently.

But what I suspect with AI is there's no such thing.

It's just localized misinformation that happens to be aggregated in a way that ultimately produces nonsense.

Think about it in medical applications.

No, I know.

But this is an early version though, right?

Yeah.

The other element that I find so interesting here is that these two advances don't come from the largest and richest organizations in the economy.

In so many ways, you always think, well, AI is sort of a game for the big guys who have unlimited access to data, who can...

throw thousands and thousands of research scientists at a particular problem.

And here you have a modest size outfit that basically does amazing work.