Two more thoughts on LLMs
- LLMs are about text. Text is back in style. We now use natural language to communicate with computers, and computers do the same to communicate back to us. A lot of graphical UI is rendered unnecessary, and the written word swells in importance.
- Near-zero vendor lock-in. Because LLMs are based in natural language, and integrating with their APIs is trivial, switching from one to the other can be done at the drop of a hat. The risk of building on top of a specific LLM is small; any good LLM will understand the prompts you used with another one. The ratio of lock in to power is orders of magnitude lower than what you’d get from a programming language or a cloud service.