banner

Overview

This week, DeepSeek, a Chinese artificial intelligence company, made significant waves in the tech industry with the release of its latest AI model, DeepSeek-R1. The model has garnered attention for its performance and the broader implications of its launch. Billions of dollars were wiped off the value of several prominent US companies reflecting weaker market confidence in future US superiority in the AI space.

Key Highlights:

Performance and Efficiency: DeepSeek-R1 is reported to perform on par with OpenAI’s o1 model, excelling in tasks such as mathematics, coding, and natural language reasoning. Notably, it achieves this performance at a fraction of the cost, being (allegedly) 20 to 50 times cheaper to use depending on the task. Subsequently sources have alleged that Deepseek may have “piggybacked” its training on the back of existing ChatGPT models (a process known as distilling) which, if true, may mean the actual efficiency advantage of deepseek may be more modest than initially claimed.

Open-Source Approach: The model is fully open-source and licensed under the MIT license, allowing for free distribution and commercialisation. This transparency contrasts with the more closed approaches of some Western AI firms making it very attractive for handling sensitive data in-house.

Market Impact: The release of DeepSeek-R1 has had a significant impact on global tech stocks. Companies like Nvidia experienced substantial stock declines, leading to a broader selloff in the tech sector. Markets have rebounded to some extent in subsequently but the impact of deepseek was both substantial and clearly unexpected.

Industry Reactions: In response to DeepSeek’s advancements, OpenAI’s responses have been mixed: CEO Sam Altman pledged to accelerate product releases (which, in of itself, may be a source of concern for some AI safety pundits) and also to improve model quality to maintain competitiveness. This may hint at the earlier release of Agents (which couple the “hands off” text/reasoning abilities of current model to add “hands on” control: taking actions and controlling both physical and electronic resources.

The timing and nature of deepseek’s emergence underscores both the rapid advancements in AI technology and is simultaneously a very clear and pointed  message from China to the US that global competition in the AI field is just starting.