Aside from concerns about the social and quality impacts of using LLM, I think using LLMs is risky. I don’t see LLMs a useful technology. However. even if I am wrong, I think it makes sense to avoid using them right now.

In a hypothetical world where LLMs turn out to be a useful technology, I see several issues with trying to use them.

1. Rapid Change

LLM technology is rapidly changing, and a high rate of change will continue as businesses invest more money into updating the tools. As a result, the functionality of LLM tools such as Claude or ChatGPT will continue to evolve and skills with using one tool may not translate. The impact of these updates will also cause significant changes to the economy.

Due to the rapid changes, I think it will sense to develop skills outside of LLM use that are more likely to remain relevant.

2. Ensuring that technology has a positive impact

As LLMs become more powerful, their ability to cause significant negative and positive impact will increase. Mitigating these impacts will require people to maintain high-trust societies where people work together to develop and test LLMs. Building trust requires people to get to know each other without LLM chatbots getting in the way.

3. Automation of knowledge work

If LLMs continue to become more powerful, they will automate a significant portion of knowledge work, which includes LLM assisted work. As a result, knowledge work will become a commodity, and skills that don’t require an LLM will become more valuable.