This is an amazing interview! A rare glimpse behind the curtain for Chinese AI. Deepseek isn't alone though, Alibaba's Qwen is actually also quite good.
I think too many people refuse to admit when they're wrong. I wasn't precisely wrong (there was nuance in the view), but I have stated, including in my interview on ChinaTalk, that I thought China would be lagging for a while. That was in October 2023, which is over a year ago (a lot of time for AI!), but I think it's worth reflecting on why I thought that and what's changed as well.
> Now, why has the Chinese AI ecosystem as a whole, not just in terms of LLMs, not been progressing as fast? This is speculation, but I’ve heard that China has much more stringent regulations on what you’re supposed to check and what the model is supposed to do. Putting that much time and energy into compliance is a big burden. A lot of Chinese tech companies and entrepreneurs don’t seem the most motivated to create huge, impressive, globally dominant models.
I never thought that Chinese entrepreneurs/engineers didn't have the capability of catching up. LLMs weren't "hitting a wall" at the time or (less hysterically) leveling off, but catching up to what was known possible wasn't an endeavor that's as hard as doing it the first time.
There's a lot more regulatory clarity, but it's actually fascinating that the culture has also shifted since then. I don't think you would have Liang Wenfeng's type of quotes that the goal is AGI, and they are hiring people who are interested in doing hard things above the money—that was much more part of the culture of Silicon Valley, where the money is kind of expected to come from doing hard things, so it doesn't have to be stated either.
That all being said, LLMs are still struggling to monetize (relative to their cost of both training and running). We'll see if OpenAI justifies its $157B valuation and how many takers they have for their $2k/month subscriptions. This is not really the sector that I would personally bet on creating a huge amount of global leadership in AI in-and-of-itself...
Except for helping train individuals and create an ecosystem where there's a lot of AI talent that can go elsewhere to create the AI applications that will actually generate value. Or be highly valuable in, say, military applications.
Anyway, again, this is an amazing interview and really suggests a big shift in the AI talent ecosystem.
I just wonder on the ways High Flyer makes it money to fund this?
This type set up needs to be done here in the USA too. Maybe have a Sovereign Wealth Fund set up to fund this type innovation? And many other future technologies too.
This is an amazing interview! A rare glimpse behind the curtain for Chinese AI. Deepseek isn't alone though, Alibaba's Qwen is actually also quite good.
I think too many people refuse to admit when they're wrong. I wasn't precisely wrong (there was nuance in the view), but I have stated, including in my interview on ChinaTalk, that I thought China would be lagging for a while. That was in October 2023, which is over a year ago (a lot of time for AI!), but I think it's worth reflecting on why I thought that and what's changed as well.
> Now, why has the Chinese AI ecosystem as a whole, not just in terms of LLMs, not been progressing as fast? This is speculation, but I’ve heard that China has much more stringent regulations on what you’re supposed to check and what the model is supposed to do. Putting that much time and energy into compliance is a big burden. A lot of Chinese tech companies and entrepreneurs don’t seem the most motivated to create huge, impressive, globally dominant models.
I never thought that Chinese entrepreneurs/engineers didn't have the capability of catching up. LLMs weren't "hitting a wall" at the time or (less hysterically) leveling off, but catching up to what was known possible wasn't an endeavor that's as hard as doing it the first time.
There's a lot more regulatory clarity, but it's actually fascinating that the culture has also shifted since then. I don't think you would have Liang Wenfeng's type of quotes that the goal is AGI, and they are hiring people who are interested in doing hard things above the money—that was much more part of the culture of Silicon Valley, where the money is kind of expected to come from doing hard things, so it doesn't have to be stated either.
That all being said, LLMs are still struggling to monetize (relative to their cost of both training and running). We'll see if OpenAI justifies its $157B valuation and how many takers they have for their $2k/month subscriptions. This is not really the sector that I would personally bet on creating a huge amount of global leadership in AI in-and-of-itself...
Except for helping train individuals and create an ecosystem where there's a lot of AI talent that can go elsewhere to create the AI applications that will actually generate value. Or be highly valuable in, say, military applications.
Anyway, again, this is an amazing interview and really suggests a big shift in the AI talent ecosystem.
I just wonder on the ways High Flyer makes it money to fund this?
This type set up needs to be done here in the USA too. Maybe have a Sovereign Wealth Fund set up to fund this type innovation? And many other future technologies too.