Uh, now in a story that was published by Rest of World last month, you actually had a quote in there which said, "AI talent is both scarce and unusually scalable." Could you tell me what scalable means in, I guess, very mechanical terms? Is that, you know, how one engineer's output basically multiplies through tooling? Or is it that the skills needed for these jobs can be acquired faster than before? I ask because, you know, these two things seem to imply opposite futures for China, so I wonder, you know, how you were kind of, uh, approaching this when you were thinking about it.

So there's an apparent contradiction that actually derives from, uh, let's say, the difference between the quantity of talent and the economic impact of talent.

Yeah, so we, we talk a lot about scarcity from a labor economics perspective.

Yeah, uh, it is scarce because, uh, you know, the demand for a very, very specific type of worker, uh, in terms of, you know, uh, top AI researchers and whose, uh-- are-- who are capable of improving model performance, and the demand has grown much faster than the supply.

Yeah, so we're not talking about software engineers in general because China, you know, produces millions of engineering graduates every year, and there, there are many people who can write code, build applications, or maybe integrate ex-existing models.

But the number of, uh, people who can train frontier models, who can improve algorithms, uh, algorithms, uh, or who can design large scale AI systems or make fundamental research breakthroughs, that pool is much smaller.

But at the same time, this talent is scalable because, uh, AI itself changes the relationship between human input and, let's say, work output or economic output.

Now, do you think this, I guess I'll call it an elite premium, compresses? Does it stay the same? Does it become even larger? And what kind of indicator would you be watching to say whether, uh, your forecasts have been correct?