New Delhi:
There is a new child on the chatbot / great language block (LLM) focused on artificial intelligence, and he threatens to blow up the rest of the water. Meet Deepseek, developed by a research laboratory based in Hangzhou with a budget fraction (if you believe reports) used to make Chatgpt, Gemini, Claude IA and others created by computer laboratories based in the United States.
And the latest offers – Deepseek V3, a parameter of 671 billion, the model “mixture of experts”; And Deepseek R1, an advanced reasoning model that uses AI, perhaps better than Openai, underlined its status as a financial and technological disturbator of potential heavy goods vehicles in this area.
How many disruptors is it?
From Monday, Deepseek V3 is the best application downloaded on the Apple Store in the United States; Let it penetrate … A Chinese developed chatbot is now the most downloaded application in the United States.
And this disturbance, even if it is considered a “potential” at that time, raised doubts about how certain American technological companies have invested the billions promised for the development of AI.
Read | Deepseek calls into question the expenses of a billion dollars in Big Tech
Anyway, the quality and profitability of the Deepseek models have overturned this story; Even if, in the long term, this particular Chinese model moves, that it has been developed with a fraction of the financial and technological resources available for Western companies is a revelation.
Again, how many disruptors is it?
Well, last month, the creators of Deepseek said The formation of the V3 model required less than $ 6 million (Although criticisms say that adding the costs of the previous development stages could push possible costs north of $ 1 billion) in IT food from the NVIDIA H800 chips, a mid -range offer. “Has Deepseek really built Openai for $ 5 million? Of course not,” Bernstein analyst Stacy Rasgon said.
But break down the available finances and it becomes quite remarkable.
OPENAI 01 bills $ 15 per million entry tokens.
Deepseek R1 invoices $ 0.55 per million entry tokens.
The price, therefore, absolutely blows the competition.
And, depending on the case of final use, Deepseek would be between 20 and 50 times more affordable and efficientthat the OPENAI model 01. In fact, the results of the logical reasoning test score are astounding; Deepseek surpasses Chatgpt and Claude Ai from seven to 14%.
Dev.ToAn online popular community for software developers, said it has marked 92% by performing complex problem solving tasks, compared to 78% per GPT-4.
By the way, the entry tokens refer to information units in the context of an prompt or a question. These are essentially what the model needs to analyze or understand the context of a request or an instruction.
For the context, Openai would spend $ 5 billion each year to develop its models.
Thus, even if the criticisms of Deepseek (see above) are right, it is always a fraction of the costs of Openai.
This is reflected, as the boss of the company pointed out, Sam Altman, in considerably improved IT capacities, but so that the Deepseek model can provide at least as much processing power on its relatively moving budget is a washing of eyebrows.
And Mr. Altman recognized it, calling the R1 model “very impressive”.
The boss of Google Sundar Pichai went further, telling CNBC in Davos: “I think we should withdraw the development of China very seriously.” And US President Donald Trump rang a “alarm clock”.
And there are the hundreds of billions of dollars that American companies have lost in the middle of a rout this week in technological actions; The NVIDIA flea manufacturer, for example, has lost more than $ 600 billion and the Nasdaq index rich in technology ended on Monday more than three%, with the independent possibility of a new drop based on the expected profits expected from Meta and Microsoft.
Read | Nvidia loses nearly $ 600 billion in Deepseek Jolts Tech
For the context, Meta and Microsoft both have their own AI models, the foreground of which are lama and co -pilot; The first is an LLM that was published for the first time in February 2023 and the second is now an integrated feature in various Microsoft 365 applications, such as MS Word and Excel.
Although it is, no doubt, at the same technological level as Openai or Chatgpt, Meta and MS have invested billions in AI and LLM projects, in the United States and abroad. For example, some analysts believe that large American cloud companies will spend $ 250 billion this year in AI infrastructure.
But what makes Deepseek special is more than cost and technology.
It is that, unlike his competitors, he is really open-source.
THE The R1 code is completely open to the public under the MIT licensewhich is a permissive software license that allows users to use, modify and distribute software with some restrictions.
This means that you can download it, use it commercially free of charge, modify its architecture and integrate it into one of your existing systems.
Deepseek is also faster than GPT 4, more practical and, according to many experts, even understands regional idioms and cultural contexts better than its Western counterparts.
There is much more consideration.
How, for example, Deepseek affects diplomatic and military links between China and the United States (and India too, in fact), and what are the ethical problems with models of AI at open source ?
But what is undeniable is that the depth of China is a disruptor. And experts believe that China has now jumped 18 to six months behind the advanced AI models developed in the United States.
Meanwhile, Deepseek’s success has already been noticed in the main Chinese political circles.
On January 20, the day he was released to the public (and the day Trump swore as president of the United States), the founder Liang Wenfeng attended a closed-door symposium for the man of Business and experts organized by Chinese Prime Minister Li Qiang. His presence was considered a sign that Deepseek could be important for Beijing’s political objective to achieve self -sufficiency in strategic industries such as AI.
With the contribution of agencies
NDTV is now available on WhatsApp channels. Click on the link To get all the latest NDTV updates on your cat.
