Hello everybody. Firstly welcome to XiaoHongShu and now welcome to DeepSeek. This is China’s Hong Pow gift of DeepSeek to the world. China is a blessing to the world. Hope these america politicians can hear and speak like you. Let us work, prosper and be happy together in this transient world. Wishing you all a Happy New Spring in The Year Of The Wood Snake
@@jsc3417 When you ask DeepSeek questions and it use inference to figure out answers for you. Once DeepSeek is released, inference engines are required running round the clock. Training are conducted massively on massive collection of data, then periodically to gather new knowlege from new data.
DeepSeek used Nvidia A80 to achieve this benchmark performance, Huawei GPU is in par with Nvidia A100. Yeaa, If DeepSeek uses Huawei' GPU to prove it can do the same or even better, US GPU embargo will totally collapse. Has anyone still double Huawei's capability in the world ? LOL
Wall Street and the media here don't want to talk about Qwen 2.5, though happy to obsess about Deepseek, because Quen has Alibaba behind it and they don't want to put attention to a large competitor. With Deepseek there is so little capital behind it that they are happy to report it, but Qwen is being missed in the reporting despite better and also free.
❤❤🎉🎉 ❤🎉 DeepSeek is democratising access to knowledge and technology, fostering a global exchange that benefits all of humanity. By freely sharing resources, DeepSeek is breaking down barriers and empowering individuals worldwide to contribute to human development and well-being. This inclusive approach transcends the limitations of traditional capitalist models, promoting collaboration and shared progress towards a more equitable and prosperous future for everyone. 🎉🎉🎉❤❤
When NIO shows up in the radars on the western hemisphere, shouldn't it also cause a similar impact on american ev manufacturers and the public will feel scammed by low quality and low tech and overpriced compared to NIO?
Getting tired of hearing about this 'sputnik moment'; I've heard this phrase over 100 times the last two days by everyone who has something to say. This is the greatest thing since chocolate got mixed up with peanut butter.
why is r1 with 50 times the inference compute advantage not much more performant per scaling laws than the arena llm leaderboard leaders google and openAI which it trails? Surely r1 should ace ARC and frontier math like o3 yet is o1 level?
These doubts about how many GPUs was used to train DeepSeek is a Big Tech paid Narrative 😂 Serious Machine learning engineers that did read the paper DeepSeek published are saying they built the algorithm to use as less computational power as possible... It's all there, open, for everyone to see. Their Machine Learning Theories are better, and works.... SIMPLE.
But even Elon thinks they cheated and have smuggled chips 😂 makes sense if u look at todays Tesla earnings report where they state they increased their chips investment so much
Wow! What a great program from Marcel. This is one of the most objective assessments that I have seen in the internet about Chinese AI development. Especially, his reference to Huawei Ascend AI chips focused on inference training. Great work Marcel!💪加油！ ❤👍👍👍👍👍🙏👏🔥😍❤️
-有趣的是，如果有证据表明 deepseek 侵犯了 openAI 的知识产权和服务条款，Groq 是否会继续托管 deepseek，因为担心被起诉- be interesting to see if Groq continues to host deepseek if is evidence deepseek violated openAI's intellectual property and terms of service for fear of being sued
@@호롱방Eres un siervo sin dignidad coreano de los americanos 😂 ¿No tuvisteis la ley marcial y un intento de golpe de estado protagonizado por vuestro presidente ahí en Corea del Sur hace nada? 😂 ¿Qué críticas a los demás con el barullo que tenéis ahí montado? Miraos a vosotros mismos
Hello everybody. Firstly welcome to XiaoHongShu and now welcome to DeepSeek. This is China’s Hong Pow gift of DeepSeek to the world. China is a blessing to the world. Hope these america politicians can hear and speak like you. Let us work, prosper and be happy together in this transient world. Wishing you all a Happy New Spring in The Year Of The Wood Snake
Deepseek used 2000 Nvidia chips for training but it used far cheaper Huawei chips for inference.
Yeah
what does it mean, inference with Huawei chip?
@@jsc3417 When you ask DeepSeek questions and it use inference to figure out answers for you. Once DeepSeek is released, inference engines are required running round the clock. Training are conducted massively on massive collection of data, then periodically to gather new knowlege from new data.
@@jsc3417 It means AI does not need many Nvidia chips (High End)
DeepSeek used Nvidia A80 to achieve this benchmark performance, Huawei GPU is in par with Nvidia A100. Yeaa, If DeepSeek uses Huawei' GPU to prove it can do the same or even better, US GPU embargo will totally collapse. Has anyone still double Huawei's capability in the world ? LOL
Game changer
DeepSeek uses Huawei's 910C for inference ever though it was trained on Nvidia. From day one DeepSeek has supported Huawei for training and inference.
The inference engine was powered by Huawei's Ascend chips
@@DONGXiiTheChinaOpportunity let make it more sensational and call it "Sputnik moment", LOL
Wall Street and the media here don't want to talk about Qwen 2.5, though happy to obsess about Deepseek, because Quen has Alibaba behind it and they don't want to put attention to a large competitor. With Deepseek there is so little capital behind it that they are happy to report it, but Qwen is being missed in the reporting despite better and also free.
no wonder my Alibaba stocks were rising.
100k+ subs...looking forward to the channel reaching that milestone! 🎉Marcel 🙏🏼
Haha far away! But a noble goal for us
Marcel, very nice to see high view count on your content! Very well deserved!!
I appreciate that!
❤❤🎉🎉 ❤🎉 DeepSeek is democratising access to knowledge and technology, fostering a global exchange that benefits all of humanity. By freely sharing resources, DeepSeek is breaking down barriers and empowering individuals worldwide to contribute to human development and well-being. This inclusive approach transcends the limitations of traditional capitalist models, promoting collaboration and shared progress towards a more equitable and prosperous future for everyone. 🎉🎉🎉❤❤
DeepSeek. Thank You. We LOVE You # ❤ from Malaysia
First! Thank you for your videos Marcel!
Thanks for watching and commenting!
Congrats on the record viewcount!!!
Thanks a lot. Overwhelming!
It will be interesting to see how this plays out. Looking forward to more of your videos!
cheers
"ASML CEO makes case that DeepSeek is positive for chip demand"
everyone got his narrative ;)
Thank you Marcel as always!! Very interesting topics
Glad you like them!
thanks for sharing your insights 👏🤞
My pleasure!
Thank you for this report
thanks for watching and commenting
Very nice! Thanks Marcel for sharing your thought.
thanks!
Interesting stuff
Thanks!
subbed ❤
Yay, thanks! 🙏
I am sure there is work on upgrading the CUDA framework as well...
100% absolutely
When NIO shows up in the radars on the western hemisphere, shouldn't it also cause a similar impact on american ev manufacturers and the public will feel scammed by low quality and low tech and overpriced compared to NIO?
it would only make sense
Getting tired of hearing about this 'sputnik moment'; I've heard this phrase over 100 times the last two days by everyone who has something to say. This is the greatest thing since chocolate got mixed up with peanut butter.
it seems like the wrong analogy anyway. Sputnik was the first satellite. DeepSeek isn't the first AI.
why is r1 with 50 times the inference compute advantage not much more performant per scaling laws than the arena llm leaderboard leaders google and openAI which it trails? Surely r1 should ace ARC and frontier math like o3 yet is o1 level?
These doubts about how many GPUs was used to train DeepSeek is a Big Tech paid Narrative 😂
Serious Machine learning engineers that did read the paper DeepSeek published are saying they built the algorithm to use as less computational power as possible... It's all there, open, for everyone to see.
Their Machine Learning Theories are better, and works.... SIMPLE.
But even Elon thinks they cheated and have smuggled chips 😂 makes sense if u look at todays Tesla earnings report where they state they increased their chips investment so much
Wow! What a great program from Marcel. This is one of the most objective assessments that I have seen in the internet about Chinese AI development. Especially, his reference to Huawei Ascend AI chips focused on inference training. Great work Marcel!💪加油！
❤👍👍👍👍👍🙏👏🔥😍❤️
🙏
-有趣的是，如果有证据表明 deepseek 侵犯了 openAI 的知识产权和服务条款，Groq 是否会继续托管 deepseek，因为担心被起诉-
be interesting to see if Groq continues to host deepseek if is evidence deepseek violated openAI's intellectual property and terms of service for fear of being sued
Deepseek use 007 chip
자유민주주의 안보 인권을 위해 딥시크 금지 제재 해야된다 자유민주주의만세
Lol aren’t you guys in Martial Law?
@@DONGXiiTheChinaOpportunity 당신혹시 중국에서오래살았나요 아님 당신러시아사람입니까..? 자유민주주를부정하는사람은 공산당 입니다
LOL! I don’t trust American billionaires either so - I don’t mind having more competition!
@ 님러시아사람인가요..?
@@호롱방Eres un siervo sin dignidad coreano de los americanos 😂
¿No tuvisteis la ley marcial y un intento de golpe de estado protagonizado por vuestro presidente ahí en Corea del Sur hace nada? 😂 ¿Qué críticas a los demás con el barullo que tenéis ahí montado? Miraos a vosotros mismos