[maxglute]

IMO it was exceptionally obvious AGI isn't around the corner because China is doing long-term AI strategy vs US market driven speculative all out sprint. When the country that produces plurality of global AI talent, including top %s do not have AFAIK any notable indications that they're hammering/evangelizing AGI like it's around the corner, it's probably not around the corner. There's also just very few Chinese AI researchers wanking about AGI timelines. There's enough AI talent there that if AGI was imminent they'd rally CCP to pursue whole of state effort, much larger than what't they're pursuing now, i.e. National Natural Science Foundation of China guideline for AI research for 2025 allocates like 15-20m USD for ~20 projects over the next 3-4 years. That's basically couch change.

[kragen]

That just means Xi doesn't believe AGI isn't around the corner. I think Xi is correct on this point, due to having watched Carmack's talk https://www.youtube.com/watch?v=4epAfU1FCuQ last month, but Xi is certainly a fallible human being, and he's not even an AI researcher, so he's even more fallible than other people whose opinions you could consult.

Here's my summary of Carmack's talk from my bookmarks file:

> #video of John Carmack saying video games, even Atari, are nowhere close to being solved by #neural-networks yet, which is why he’s doing Keen Technologies. They’re going to open-source their RL agent! Which they’re demoing playing an Atari 2600 (emulated on a Raspberry Pi) with an Atari 2600 joystick and servos (“Robotroller”), running on just a gamer laptop with a 4090. (No video of the demo, though, just talking head and slides.) He says now he’s using PyTorch just like anyone else instead of implementing his own matrix multiplies. Mentions that the Atari screen is 160×210, which I’d forgotten. They’re using April-tag fiducials so their camera can find the screen, but patch them into the video feed instead of trying to get the lighting right for stickers. He says the motion-to-photons #latency at Oculus had to be below 20ms for avoiding nausea (26'58”). Finding the Atari scores on the screen was surprisingly hard. #retrocomputing

[maxglute]

I think it means Xi's advisors don't believe in AGI, and Xi's advisors are drawn from the largest pool of AI talent in the world. And given the incentive, i.e if they're wrong and PRC lightcone collapse because US built AGI, they would if anything advise AGI is more imminent to hedge. But they're not. The general discourse around PRC AI doesn't focus on AGI. IMO simply one one thinks AGI is around corner, not enough to drive short term strategic planning.

[kragen]

I agree, but the ones who are drawn from that pool are not necessarily the ones who are least wrong. So far, though, I have to admit, Xi's technological predictions have been solidly on target.