Photo by Matheus Bertelli on Pexels

DeepSeek, a Chinese AI startup, is shaking up the artificial intelligence world with its resource-efficient approach. The company has developed AI models that rival those of established Silicon Valley giants, but with a fraction of the computing power. This feat is achieved through a focus on algorithmic efficiency and a commitment to open-source collaboration.

DeepSeek’s AI assistant quickly surpassed OpenAI’s ChatGPT to become the top free app in the US App Store. The company leverages innovative techniques like Multi-head Latent Attention (MLA) and Group Relative Policy Optimisation (GRPO) to overcome memory constraints and optimize reinforcement learning processes.

Remarkably, DeepSeek operates with significantly fewer GPUs than its competitors, achieving a tenfold improvement in resource utilization. This emphasis on efficiency, combined with its open-source philosophy (in contrast to the often secretive practices of major US tech companies), has garnered attention and praise, including from prominent figures like Marc Andreessen.

DeepSeek’s rise suggests a potential paradigm shift in AI development, moving towards more accessible and sustainable methods. The company’s impact is already prompting a reassessment of existing, resource-heavy practices within the broader AI industry.