Zuckerberg warns: Deepseek and other Chinese AI models are too powerful, US companies and government should do everything possible to block them.

The AI model published by Chinese startup DeepSeek has recently exploded and today topped the free list of Apple's iOS App Store in the United States, which has caused Meta founder Zuckerberg to be jealous by creating a model that can compete with ChatGPT and Meta Llana for less than $6 million. (Synopsis: ChatGPT o3-mini coming soon!) OpenAI product long plus: Q1 launches AI Agents tool) (background supplement: OpenAI released a full-blood version of the o1 model and a new subscription scheme ChatGPT Pro, is the monthly fee of $200 worth it?) The day after Christmas last year, DeepSeek, a Chinese Hangzhou-based artificial intelligence lab, launched DeepSeek-V3, a free Open Source large-scale language model, which shocked the global technology circle, became an overnight hit and became the focus of the recent market. Why exactly has DeepSeek caused fear among Silicon Valley AI giants Microsoft, Meta and others? Simply put, it is because it is said to have spent less than $6 million, using Nvidia's low-performance H800 chip, and it took only 2 months to successfully create a DeepSeek-V3 that matches or even exceeds the most powerful model on the market in the United States. DeepSeek spends only $5.6 million to surpass ChatGPT, Llama model According to CNBC, in a set of third-party benchmark tests, DeepSeek's model accuracy performance is better than OpenAI's GPT-4o, Meta's Llama 3.1, and Anthropic's Claude Sonnet 3.5, according to CNBC. Less than 1 month, on January 20, DeepSeek's newly launched "DeepSeek-R1" inference model also outperformed OpenAI's latest o1 model in many third-party tests. Chatbot Arena, a well-known AI system ranking website, shows that "DeepSeek-R1" ranks third on the overall list, and ranks first with the top inference model o1 launched by OpenAI. DeepSeek-V3 ranks eighth overall, topping the Open Source model and the most cost-effective model in the top ten. DeepSeek-V3 only cost less than $5.6 million to build, compared with OpenAI, Google, Microsoft and other companies every year billions or even tens of billions of dollars in training costs, it can be said that it has achieved a leap in cost-effectiveness, triggering a panic among AI companies in Silicon Valley. U.S. tech stocks suffered, Huida fell more than 3% on Friday DeepSeek is also a wake-up call about whether the U.S. global leadership in AI is being caught up by China and calls into question the hundreds of billions of dollars big tech companies spend on building AI models and data centers. Some people in the market even attributed the reason for Huida's big dump price of 3.12% on Friday to the explosion of DeepSeek, believing that DeepSeek may be the biggest threat to U.S. stocks. Bloomberg reported today that DeepSeek has raised questions about U.S. tech dominance. U.S. stock index futures tumbled heavily in early Asian trading on Monday. Extended reading: OpenAI, SoftBank, Oracle launch the largest AI infrastructure plan in history "Stargate Project": Invest $500 billion to set up a data center in the United States DeepSeek caused panic within Meta, Zuckerberg warned: The AI gap between China and the United States is very small "Information" broke the news that DeepSeek's rapid progress has put Meta's AI team into panic mode and worried about Meta's flagship Llama The next version of the model may not be able to compete with DeepSeek, lag behind the AI race, and even rumor that Meta has set up 4 war chambers to study how DeepSeek works. Researchers at other top AI companies, such as OpenAI, are also analyzing the DeepSeek model in an effort to find insights in terms of cost efficiency. Meta CEO Zuckerberg also recently publicly named DeepSeek as a very advanced model, "the AI race is a global geopolitical war, and now the gap between China and the United States is very small", and called on the US government not to hold back, give more support to American companies, and do not take the current leading position for granted. Microsoft CEO Nadella (Satya Nadella) was also impressed by DeepSeek's new model, praising them for really effectively completing an Open Source model, doing inference time calculations, and having supercomputing efficiency, "We should take developments in China very, very seriously." Midjourney founder praised DeepSeek for strong Chinese and English skills AI drawing tools The founder of Midjourney recently praised DeepSeek's excellent Chinese and English skills, and he tweeted yesterday: Deepseek has crushed Western models in ancient Chinese philosophy and literature, and his mastery of English is much better than my first-hand Chinese sources. It felt like a cross-generational exchange of literary/historical/philosophical knowledge that I had never been exposed to before, which was quite touching and justified. Western labs don't care about training on Chinese data (but Chinese labs train both). Keep in mind that China has a literary history of thousands of years longer than the West (because we lost most of our Roman/Greek/Egyptian literature, while China retained theirs). Basically, our model lacks the literary basis of Western thought, but the Chinese model is intact. This may be either a "data advantage" or a less obvious advantage in spiritual and philosophical self-actualization. It is worth noting that DeepSeek is also rapidly attracting a large number of users, and its app has surpassed ChatGPT in the United States with the top of Apple's app store free download rankings in the United States and China, which is a major initiative and has become another Chinese phenomenon after Tiktok that has exploded in the United States. DEEPSEEK OVERTAKES CHATGPT IN APP STORE RANKINGS IN THE U.S. Chinese AI platform DeepSeek has reportedly surpassed OpenAI’s ChatGPT on Apple’s App Store rankings just a week after launch. Developed by Hangzhou-based DeepSeek, the platform offers advanced reasoni...