What Sarah Guo, Scott Kennedy, and others miss is that it's the sanctions that went into effect in mid-November, less than two months ago, that should have a real effect. Previous sanctions were simply annoyances. This being said, it's quite possible DeepSeek, Alibaba, and others could copy o3's secret sauce and offer it at a much more reasonable price.
However, it's not o3 that's going to get us to level 5 AI or ASI.
Expect to see a lot of great models coming out of China this year. Next year, maybe not so much. And it's next year -- and level 5 AI and ASI -- that matters. o3 will seem like a toy in comparison.
My browser takes 500mB of RAM to play a YouTube video. It should surprise no one that China is responding to hardware restraints by picking up ready wins in software efficiency. Western/Indian coding is incredibly inefficient.
Anyway, thanks China, I don't want my oligarchs dominating the LLM space.
"A quick point: this notion going around about “winning the AI war against China” is complete and utter bullshit. It’s just more propaganda to justify a war with China.
So is the “democracy vs authoritarianism” debate over AI. I’m an anarchist - so distinguishing between the Western version of “authoritarianism” and “Chinese (or Russian or Iranian) authoritarianism” is a waste of my time.
ALL states are “authoritarian” - they just differ in degree. And they all get worse over time."
To continue that point, it doesn't matter who achieves "AGI" - whatever the hell that it, since there is no precise definition, just the vague "it can do anything", which is complete bullshit from a scientific viewpoint.
In addition, anyone who believes the utter crap coming out of the mouth of Sam Altman is a moron. He's in it for the money and don't let anyone tell you different. So he's going to run his mouth about how his company's product "can do anything" and only suckers who don't know the difference between "promotion" and "lying" - there is none - will fall for it.
And as to who is going to use AI - of any level - for "authoritarianism": Who put the former head of the NSA on his openAI's Board of Directors? If you think the NSA is here to "defend America" from China or anyone else, I have a bridge to sell you.
China is not a threat to the world, AI or no AI. Neither is Russia, Iran or any of the other "threats du jour". The US is the one with 800 military bases across the entire world and the NSA spying on literally everyone on the planet, including every post on social media, including this one. It is the US that spends five times more than any other major country combined on military technology for the profit of the military-industrial-espionage complex.
For followers, open-sourcing LLMs lets the AI community study and innovate, which is more valuable than keeping them closed. It’s a smart move that also boosts their tech brand. Plus, many still use DeepSeek APIs (so cheap) as running an MoE 671B model (at scale) requires expertise.
What Sarah Guo, Scott Kennedy, and others miss is that it's the sanctions that went into effect in mid-November, less than two months ago, that should have a real effect. Previous sanctions were simply annoyances. This being said, it's quite possible DeepSeek, Alibaba, and others could copy o3's secret sauce and offer it at a much more reasonable price.
However, it's not o3 that's going to get us to level 5 AI or ASI.
Expect to see a lot of great models coming out of China this year. Next year, maybe not so much. And it's next year -- and level 5 AI and ASI -- that matters. o3 will seem like a toy in comparison.
My browser takes 500mB of RAM to play a YouTube video. It should surprise no one that China is responding to hardware restraints by picking up ready wins in software efficiency. Western/Indian coding is incredibly inefficient.
Anyway, thanks China, I don't want my oligarchs dominating the LLM space.
I recently said this on a Note:
"A quick point: this notion going around about “winning the AI war against China” is complete and utter bullshit. It’s just more propaganda to justify a war with China.
So is the “democracy vs authoritarianism” debate over AI. I’m an anarchist - so distinguishing between the Western version of “authoritarianism” and “Chinese (or Russian or Iranian) authoritarianism” is a waste of my time.
ALL states are “authoritarian” - they just differ in degree. And they all get worse over time."
To continue that point, it doesn't matter who achieves "AGI" - whatever the hell that it, since there is no precise definition, just the vague "it can do anything", which is complete bullshit from a scientific viewpoint.
In addition, anyone who believes the utter crap coming out of the mouth of Sam Altman is a moron. He's in it for the money and don't let anyone tell you different. So he's going to run his mouth about how his company's product "can do anything" and only suckers who don't know the difference between "promotion" and "lying" - there is none - will fall for it.
And as to who is going to use AI - of any level - for "authoritarianism": Who put the former head of the NSA on his openAI's Board of Directors? If you think the NSA is here to "defend America" from China or anyone else, I have a bridge to sell you.
China is not a threat to the world, AI or no AI. Neither is Russia, Iran or any of the other "threats du jour". The US is the one with 800 military bases across the entire world and the NSA spying on literally everyone on the planet, including every post on social media, including this one. It is the US that spends five times more than any other major country combined on military technology for the profit of the military-industrial-espionage complex.
'
Get a fucking clue.
What do you think are the incentives for releasing Deepseek model weights for free? Hedge fund companies are not known for their ... generosity.
For followers, open-sourcing LLMs lets the AI community study and innovate, which is more valuable than keeping them closed. It’s a smart move that also boosts their tech brand. Plus, many still use DeepSeek APIs (so cheap) as running an MoE 671B model (at scale) requires expertise.