How is China's new language model better?
DeepSeek has trained its model using GPUs that are considered 'last generation' in the US. DeepSeek's results are competing with expensive models like GPT-4 and Meta's Llama. DeepSeek-V3 has won the race. It has left behind big models like OpenAI's GPT-4, Anthropic's Cloud 3.5 Sonnet and Meta's Llama 3.3. Be it coding, solving math problems or finding errors in code, DeepSeek-V3 is at the forefront.
And just as the AI world was discussing DeepSeek-V3, DeepSeek launched another new model last week, DeepSeek-R1. This R1 model has even surpassed OpenAI's new O1 model.
Why is there alarm bells for America?
DeepSeek is taking over the world and this is happening at a time when OpenAI is trying to make itself a money-making company. DeepSeek has made its AI models open-source, meaning anyone can use and change them. This is completely different from OpenAI, which is why DeepSeek is having an even greater impact on the world.
Being open-source means that developers from around the world can further improve DeepSeek's models. This will allow China's AI technology to become a part of the world's tech system, which only the US has been able to do so far.
If China's AI technology starts being used in phones, computers and other gadgets around the world, it will be a warning bell for the US. DeepSeek has opened a path through which China's AI can take over the world.
click and follow Indiaherald WhatsApp channel