A Chinese Startup's New AI Model, DeepSeek R1, Shakes the AI World

DeepSeek AI: The Chinese AI startup has the tech world talking after releasing two large language models (LLMs) to rival dominant US-developed AI tools. The Liang Wenfeng-founded company says the free AI assistant utilizes less data and computer power and is built at a fraction of the cost of other advanced AI tools.

Despite US export controls on chips and critical hardware, DeepSeek has made strides in AI with its resource-efficient open-source models. 

Evidence shows that the sanctions aren’t weakening China's AI capabilities. They drive Chinese startups to innovate harder, prioritizing efficiency, collaboration, and top talent pooling.

DeepSeek-R1, an AI model released in January, has the same reasoning capabilities competing with advanced foundation models but is cheaper to operate than OpenAI because it utilizes fewer advanced chips. This possibility led to the most significant one-day loss in US history last month, with NVIDIA shedding almost $600bn in market value.

The reasoning model produces incremental responses, much like humans' reasoning through ideas and problems, using less memory than OpenAI. DeepSeek, like other Chinese AI models like ByteDance’s Doubao and Baidu’s Ernie, is trained to avoid politically sensitive questions.

DeepSeek-R1 fits the Chinese government’s ambition to be the global leader in artificial intelligence. Its success undoubtedly reshapes the AI world and demonstrates China’s significant influence in the artificial intelligence and tech industry.

R1 is an open-source language model with an MIT license, allowing free commercial and academic use. Much like its direct competitor ChatGPT, R1 can perform text-based tasks and solve scientific problems. 

The Chinese startup also launched another model, Janus-Pro-7B, which can generate images from a text prompt like OpenAI’s DALL-E 3 and Stable Infusion’s Stability AI. 

AI companies like DeepSeek will inevitably emerge in China for several reasons: a growing pool of talented people with doctorates in technology, science, mathematics, and engineering, as well as colossal venture capital investments in LLM-developing firms. 

Follow Puppetry for more AI news!