Small Language Models will dominate the revenue earning AI applications. Specific databases for specific purposes, such as writing legal contracts, researching medical issues and other research topics. An LLM isn’t effective for these applications with trillions of data units to sort through rather than the SLM which has only the relevant 10bn to 15bn data bits, and runs on a desktop computer.
LLMs will only be for limited commercial applications, and the internet dross of people using it as a super google search engine, or as a companion. Which won’t pay. Massive mis-allocation of capital will crash the AI and chip stocks, and nuclear power companies.
