What happens if LLM’s suddenly change their cost to be 1000 USD per user per month? What if it is 1000 USD per request? Will new students and new professionals still be able to complete their jobs?
What happens if LLM’s suddenly change their cost to be 1000 USD per user per month? What if it is 1000 USD per request? Will new students and new professionals still be able to complete their jobs?
Calculators have never been more accessible/available. (And yet I personally still do most basic calculations in my head)
So I agree students should learn to do this stuff without LLMs, but not because the LLMs are going to get less accessible. There's another better reason I'm just not sure how to articulate it yet. Something to do with integrated information and how thinking works.
LLM’s are not consistent. For example, having a new company make a functional duplicate of ChatGPT is nearly impossible.
Furthermore, the cost of LLM’s can change at any time for any reason. Access can be changed by new government regulations, and private organizations can chose to suspend or revoke access to their LLM due to changes in local laws.
All of this makes dependence on an LLM a risk for any professionals. The only way these would be mitigated is by an open source, freely available LLM that creates consistent results that students can learn how to use.
I could be wrong, time will tell. I just wouldn't base my argument for why students should learn to think for themselves on accessibility of LLMs. I think there's something far more fundamental and important, I just don't know how to say it yet.
LLMs are becoming increasingly efficient. Through techniques such as distillation, quantization, and optimized architectures, it is already possible to run capable models offline, including on personal computers and even smartphones. This trend reduces reliance on constant access to centralized providers and enables local, self-contained usage.
Rather than avoiding LLMs, the rational response is to build local, portable, and open alternatives in parallel. The natural trajectory of LLMs points toward smaller, more efficient, and locally executable models, mirroring the path that calculators themselves once followed.