美国工业安全局老大Estevez临别坦言:老中工程师“smart”

z
zavanna
楼主 (北美华人网)

Estevez conceded that US policy could do no more than slow China since it had smart engineers, money to invest and government support. 
But he said Washington had been successful, pointing to statements from Chinese AI groups identifying US export controls as their main obstacle. “We’ve definitely impeded the People’s Republic of China’s ability to produce those chips and ergo have impacted their ability to develop large language models,” said Estevez, referring to the models that power AI products.
f
feifeiduan
他指的是不卖最先进AI芯片给中国吧?
托美国的福,有了Deepseek
s
sayuki
Estevez conceded that US policy could do no more than slow China since it had smart engineers, money to invest and government support. 
But he said Washington had been successful, pointing to statements from Chinese AI groups identifying US export controls as their main obstacle. “We’ve definitely impeded the People’s Republic of China’s ability to produce those chips and ergo have impacted their ability to develop large language models,” said Estevez, referring to the models that power AI products.
风轻雨
sayuki 发表于 2025-01-21 20:49
Estevez conceded that US policy could do no more than slow China since it had smart engineers, money to invest and government support. 
But he said Washington had been successful, pointing to statements from Chinese AI groups identifying US export controls as their main obstacle. “We’ve definitely impeded the People’s Republic of China’s ability to produce those chips and ergo have impacted their ability to develop large language models,” said Estevez, referring to the models that power AI products.

那就大家公开竞争呗,比得就是谁聪明,谁努力。
美国玩这种招,丢人啊。
s
summerline
事实就是中国还是做出来了deepseek而且开源,把细节都发布了。 看看现在ai的文章,有几篇是没有华人名字的。 所以老黄都去中国了。
a
amyonego
回复 1楼 zavanna 的帖子
US policy could do no more than slow China
就是承认挡不住,只能拖延几年?
a
amyonego
summerline 发表于 2025-01-22 16:42
事实就是中国还是做出来了deepseek而且开源,把细节都发布了。 看看现在ai的文章,有几篇是没有华人名字的。 所以老黄都去中国了。

deepseek开源了吗?
m
minqidev
事实就是中国还是做出来了deepseek而且开源,把细节都发布了。 看看现在ai的文章,有几篇是没有华人名字的。 所以老黄都去中国了。
summerline 发表于 2025-01-22 16:42

今天出了一个R1版本,基本上,中国有自己发展道路的开源类似open ai的东西了。open ai不开源,反倒是 deepseek开源了。 ( ̄▽ ̄)",开源有人实现了100线程,20万字输出。。 这个其实太牛了,等于说以后可以一次性,写书了
m
minqidev
deepseek开源了吗?
amyonego 发表于 2025-01-22 16:45

你才上网的?
GitHub 组织主页 https://github.com/deepseek-ai 这里包含了多个开源项目,如数学推理模型、对话模型等。
Hugging Face 模型库 https://huggingface.co/deepseek-ai 可以直接访问预训练模型和说明文档。
主要开源模型示例: DeepSeek-Math(数学推理模型) GitHub: https://github.com/deepseek-ai/DeepSeek-Math
DeepSeek-R1(对话模型微调优化框架) GitHub: https://github.com/deepseek-ai/DeepSeek-R1