China Morning Missive
Interesting to see that the latest iteration of the Deepseek AI model isn’t getting much attention in the business media. Thankfully there’s always Tom’s Hardware.
There is trend becoming increasingly apparent. Given the ongoing restrictions placed on China by the American government, AI developers are having to create unique solutions to solve the issue of compute. Here we have just one example taken from the linked article.
“Chinese developers of Deepseek AI have released a new model that leverages its multi-modal capabilities to improve the efficiency of its handling of complex documents and large blocks of text, by converting them into images first. Vision encoders were able to take large quantities of text and convert them into images, which, when accessed later, required between seven and 20 times fewer tokens, while maintaining an impressive level of accuracy.”
Again, this is just one example and provides the sort of roadmap being used by all AI developers in China. Keep in mind as well that the Chinese models are all open sourced and, in nearly all cases, open weight as well. Iteration among the larger players, such as QWEN and Kimi K2, explains why these groups have been so aggressively quick with the release of enhanced models.
The same, too, holds for the application of these of models. In the Deepseek example, Chinese companies ranging from automotive to logistics are finding ways to use the AI models to enhance overall performance in production and/or servicing. The primary objective here is to build for scale and in doing so provide actual usability and with deep cost effectiveness.
A stark difference when compared to all the financial shenanigans taking place among the various American AI groups.
Tom
New Deepseek model drastically reduces resource usage by converting text and documents into images —
Could help cut costs and improve the efficiency of the latest AI models.
