They are better than me at many tasks.
> Not everything has to be solved by cheap* technological processes.
> *: If you ignore the environmental costs.
For many tasks, inference on an LLM is a lot cheaper (including for the environment) than keeping a human around to do them. As a baseline, humans by themselves take around 100W (just in food calories), but anyone but the poorest human also wants to consume eg housing and entertainment that consumes a lot more power than that.