Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Artificial intelligence only the moment of Sputnik.
Businessman Mark Andresson did so A bold demand for xThe previously known social media platform as Twitter last Sunday. Silicon Valley, along with the stock market and the vowed online employees, appears to be an earthquake activity in the artificial intelligence space.
Deepseek Ai, a new model of artificial intelligence from China JumpEcho is sent throughout the silicon valley. Deepseek claims to compete with artificial intelligence, and in some cases It excelsOpenai O1 thinking in a small part of the cost. Not only that, the Deepseek R1 model is completely open source, which means that the code can be accessed publicly and anyone can use it for free.
One of the main distinction between Deepseek-R1 and Openai’s O1 is that R1 allows you to see its series of ideas. It is an incredible vision on how to “think” of artificial intelligence. “You can actually see it Try to answer questions about Tiananmen Square Before canceling its response, in the middle of the road. Nafidia, the company that makes the chips that work on the revolution of artificial intelligence, saw it Arrows are drowned 18 % And a Register 600 billion dollars After the weekends of Deepseek. From logical. If what Deepseek says is correct, it is achieved near the performance of the O1 level on the old NVIDIA chips, apparently during the spending of a small percentage of the cost.
Online commentators are still trying to understand the sudden appearance of Deepseek in the artificial intelligence market. Is it actually a performance with O1 at a lower cost? To what extent can Deepseek and China claims are correct with regard to efficiency? Does costs come from a major technical opening, or are other areas of the supply chain in China make their use cheaper?
Regardless, R1 is impressive.
Mel Morris, CEO of Corpora you haveArtificial Intelligence Research Engine, in a statement of CNET. Morris added that Dibsic offers competition for artificial intelligence players, and its existence is likely to stimulate faster progress in artificial intelligence technology, which leads to more efficient and accessible solutions to meet the increasing demand.
It could be the reason that the CEO of Openai Lower For the O3 Mini Information near the end on Saturday.
Since Big Tech is constantly throwing billions of dollars, processing, energy and energy in artificial intelligence, Deepseek’s efficiency can be closer to the type of jump that we saw when cars moved from the carp to fuel injection systems. Unlike Openai, the Deepseek R1 model is an open source, which means that anyone can use technology. It is a significant disruption in the market, which is currently dominated by Chatgpt from Openai and Google’s Gemini, both closed and require users to pay to completely reach their features.
In the artificial intelligence race between the United States and China, America remained at the forefront thanks to the emptying of the massive investment in Silicon Valley and the exclusive government on NVIDIA to sell the latest artificial intelligence chips to China. However, this siege may only be Getting China to make its own chips faster. Money, in addition to protectionism, was seen as a way to maintain China in second place, which causes the world to depend on American technology. This dynamic has turned. Now, consumers and companies around the world can reach a very “thinking” model in a small part of the cost. Not only that, Tiktok has released the mother by atteedance Even the cheaper model for R1.
With markets and social media response to new developments outside China, it may be too early to say that America has been beaten. But at least, China is causing quickly.
“China has already produced GPT-4 quality models, but there was a longer late delay-just as it took a year, for a year and a half, something like this. But now there is a Chinese model, only six may have said Lucas Hansen, co-founder of CIVAIIt is a non -profit organization that uses programs to show what Amnesty International can. “Therefore, the United States still has progress, but it is not as large as it was before.”
The only thing that will help artificial intelligence companies to catch up with Openai is the R1 ability for users to read their ideas series. Even if the R1 does not get every answer correctly, the ability to see how reasons can help better develop it. “Shock and awe” with R1 comes from the ability to read her series of ideas, according to Hansen. It is Insight Openai did not allow access to its O1 model, as hiding the secret sauce keeps people screaming at the cost of a monthly subscription to reach.
However, there is a level of doubt that should be taken as a cost cost for R1 performance. the The white paper published by Dibsic had more than 100 participating authors. This is a lot of mental strength to train artificial intelligence at a low cost of $ 5.5 million. The cost of only $ 5.5 million may be energy costs to train the model, minus individual salaries for all researchers, but China was not completely transparent on how to calculate these energy costs. The cost of establishing a data center in China is likely to differ from one preparation in the United States. It is not certain whether the costs are supported by the provider of the cloud or the Chinese government itself, according to Hansen.
There are also doubts on the Deepseek chips used to train their model. It is the company in fact Using NVIDIA A100 and H800 chips older or China reaches the latest H100 chips through other meansAnd Alexander Wang, CEO of Scale Ai.
Even if we take this number of $ 5.5 million as a very conservative appreciation, it is still much less $ 100 million, cost it Openai to train GPT-4Companies former model for Amnesty International. Openai has not released numbers about the cost of building O1, but given the cost of the distinctive code much higher for customers, it is likely to be more expensive.
“With the download of the data center in the United States, it is expected to double or three times by 2030, any efficiency savings can have a significant impact,” said Mark James, the interim director of the Energy and Environment Institute in Fairmont Lu. . actually, The facilities are emphasized through high energy requirements of artificial intelligence. If Deepseek claims are correct, it may significantly reduce potential electricity load, which reduces stress on both consumers and the environment. “On the other side, the most efficient models can open more growth in this sector, which will reduce efficiency savings and exacerbate tension on our network,” said James.
It claims that the United States has It lost the war of artificial intelligence It may be too early. At least, the scene has become more competitive immediately and there is room for continuing innovation. Deepseek also does not mean that the world is on the verge of achieving Artificial general intelligenceOr advanced advanced AI that is more intelligent than humans and can learn itself.
“I don’t think Deepseek brings us closer to us one millimeter of artificial general intelligence (AGI), but I think it brings us closer to the applications of the large lifeline language model (LLM), which is great,” he said. Bin Ghawartl CEO of the AI (ASI) coalition and founder founder Singularitynet. Deepseek still has the same cognitive restrictions as other artificial intelligence models. Nevertheless, Deepseek’s efficiency can press artificial intelligence.