- After the increase in the popularity of Deepseek, Nvidia is commenting on the newcomer of AI
- Nvidia calls Deepseek an ‘excellent advance of AI’
- Nvidia suggests that their GPU is still critically important
If I had not heard the uproar about Deepseek over the weekend, there are many possibilities that at least he has heard the term for now. He jumped to fame because he provided a genuine competitor to chat at a price fraction, and has caused agitation in the stock market, seeing that the prices of technological actions collapse. Nvidia in particular suffered a record price of $ 600 billion prices, the highest drop in history prices.
Throwed by a Chinese startup of the same name, Depseek is a free chatbot with ambitions to face Operai’s chatpt tastes. There are also new models with some multimodal capacities, mainly in the creation and analysis of images. He has taken the world of AI assault and remains the number one application in Apple’s application store in the United States and the United Kingdom.
The application and the website proved to be popular, with Depseek experiencing an interruption and a ‘malicious attack’ the same day that went to fame.
Although Sam Altman, Executive Director of OpenAi, replied, we also listened to Nvidia, possibly the world leader in AI chips, who has increased in prominence as the wave of AI has continued to grow.
In a statement sent by email to Techradar, Nvidia wrote, “Depseek is an excellent advance of AI and a perfect example of the test time scale. Deepseek’s work illustrates how new models can be created using that technique, taking advantage of widely available models and calculating that it meets export control. Inference requires a significant number of NVIDIA GPU and high performance networks. Now we have three scale laws: pre-training and post-training, which continue, and the new test time scale. “
It is certainly strong, calling Depseek “an excellent advance of AI”, which speaks of the performance of the Deepseek model. It also confirms what we knew: new models can be established using existing models and chips instead of creating completely new ones.
Nvidia clearly wants to remain a key part, noting that this type of deployment requires many NVIDIA GPUs and plays the fact that Deepseek used the specific NVIDIA GPUs of China. Reading between the lines, he also suggests that Depseek will need more of his chips … at some point.
Deepseek states that he used a new innovative training process to develop his LLM using self-show evidence and errors. You could say that he trained his LLM in the same way that humans learn, upon receiving comments based on their actions. He also used a MOE architecture (mixture of experts), which means that he activates only a small fraction of his parameters at any given time, significantly reducing the computational cost, which makes it more efficient.
Sam Altman also praised the Depseek model, “particularly around what they can deliver for the price.” He reiterated that Operai “will obviously deliver much better models,” but welcomed the competition. Nvidia seems to be maintaining her future letters closer to her chest.
It is still a kind of waiting game to see when Depseek AI will turn on new records and return to complete performance, but it is curious about its power of permanence, read my colleague Lance Ulanoff’s-The editor of Techradar-A-Large – thoughts About his chances of staying in the United States. As well as Our Deepseek AI versus Chatgpt practice By John-Anthony Dysotto, one of the experts in the PakGazette.
