SHOULD BIG TECH GO FRUGAL ?
DeepSeek has actually indulged in frugal-engineering while developing, what seems to be, a good LLM based AI assistant called R1. To me, the use of fewer high-performance and AI-specific chips, like those built by Nvidia, smacks of gobs of frugality going into DeepSeek’s creation of R1. DeepSeek R1 is an advanced-frugal-product (AFI) that by using, quite literally, fewer resources and, hence lower costs, has upped the ante in the development of cost-and-resource effective LLM tools of superior quality. DeepSeek R1 also involved significant involvement of Chinese researchers for achieving this feat, which is another hallmark of an AFI, use of a workforce with advanced degrees to “frugalize” challenges in science.
Deep Seek R1 has trained its machine learning models with fewer parameters and hence, fewer chips, together with efficient data storage. In short, DeepSeek R1 is an example of frugal-computing, i.e., effective computing done affordably with lesser resources by a talented workforce for achieving top quality.
I believe companies, businesses, and other entities should engage in frugal-engineering thereby embracing artificial constraints on resources and costs! Such artificial constraints enable the best of scientific efforts to tackle even challenging problems for giving top quality performance. Constraining use of resources and costs is important to the global industry. Artificially limiting what you consume to produce, when surrounded by ample resources, not only unleashes the best of science and engineering but also helps in allocating resources globally for a better world.
Interestingly, the frugality-based business philosophy originated in India and China, with both countries now making the best use of it in achieving important milestones!
Frugal-engineering examples abound including amazing frugalized versions of cars, space-tech and even particle-accelerators, to name a few. Please read my book on frugal-engineering, which presents many examples of frugal-tech and outlines procedures to methodically create AFIs in various sectors.
https://link.springer.com/book/10.1007/978-981-99-9700-8
Please note that these are just my personal views!
Suggested reading:
