it’s getting to the point where I notice people say it a lot, especially IRL now for whatever reason recently.
And for clarity I’m not in research or anything, so these people just mean ‘LLM/image gen’, not utilities like OCR or (usually not) transcription.
Some have argued it’s just more efficient (which I can kind of get), while others think you’re actively hindering your intelligence somehow.
On the first point:
I’ve tried it occasionally to see how it compares to my own skill, and while it produces a functional result, it’s always very derivative work to the point where you can find things with the exact same names of other ‘public’ (but not libre) works, and often isn’t the ideal solution to what it targets. So I can see how you can get things out of it, but it never felt really that profound to me.
But for the second… isn’t this supposed to be the tool for people to do things they aren’t experienced in? If anything, you probably need to be able to understand how to write pertaining to the task so the token probabilities are biased toward writing from that area.
And even then, if all you end up doing is prompting AI, then wouldn’t you ultimately serve no purpose outside of being glorified QA?
I guess I’m trying to figure out what exactly non-users would be ‘falling behind’ in that affects them more than those who use AI?
A preamble you didn’t ask for (feel free to skip if you don’t wanna talk shit on my views. You’re welcome to talk shit, but I ask you at least understand my position first.) I don’t like AI because of the societal, economic, environmental, and imperialist effects of it. Actually, I despise it for that.
However, LLMs and ML overall has seen some profound benefits from these investments, and there is real, unique benefit to understanding their strengths, limitations, etc.
Nobody who says that is talking about a chatbot, and if they are they should shut the fuck up.
I will also say, I have a pretty strong moral opposition to using US products specifically. Their training is extremely inefficient, their power sources are horrible for the environment, they’re being used primarily to disenfranchise the working class rather than benefit humanity, and their data enrichment process are hugely exploitative in the best scenarios, and literal (like, literally literal) torture.
There’s truth to the utility of the models, and there are models without such grotesque wrongdoings that are much easier to get behind. China is currently, accounting for cost, absolutely neck and neck with the US in practical capability, and also requires datacenters to minimize impact to the surrounding communities as well as use renewable energy.
My response to the post: LLMs are good for programming and other work around similarly syntax-structured entities.
Yeah, it’s pretty good as a search engine for a topic you’re unfamiliar with, and getting you a functional thing. Whether software, a study method, instructions for soup, it can probably help you do it as well as you could as a novice otherwise, which much less initial effort and headaches. That’s the intuitive use without much requirement of or room for ‘excellence’ (in the most literal sense).
I think the areas people can become better than others pertain to some things you mentioned and some others; prompt efficacy, context management, estimating effort level, harness optimization, etc.
When you get into the really bleeding edge of LLM-centric development, optimization and yes, engineering the automation around prompt structure, routing, effort estimation and evaluation, contract systems, etc. can mean the difference between 30k waste over the course of a month, or a solid set of product improvements.
For the QA argument, I disagree, at least partly. The people who were never going to be great devs or architects probably will end up as glorified QA for as long as employers allow that to happen. I think, in my own and my peers’ experience at least, that AI really does allow more human time to be spent on broader scoped tasks than before.
I don’t want to make an assembler or compiler analogy because I think it’s tired and reductive, but I think it’s pretty on point here; less human time is spent on entry, and a higher ratio of human time is spent on problem solving. That wording is deliberate; I haven’t been convinced it will proliferate the work in the same way other advancements have; I don’t think it’s a given.
https://www.oecd.org/en/publications/pisa-2025-results-volume-i_73451bc5-en.html
Students who use LLMs a lot are overall much worse off then their regular counterparts. Now, whether that means slower students use LLM more (without it having any real effect) or that LLM use is degrading their ability is unknown, but it’s pretty clear and extremely damning.
I agree they are capable of that, but I think that’s improper use and is deserving of its own slur: meat proxy.
The same result has been shown for calculators, and I think it’s the same problem. Overreliance is not a new problem.
I like physical strength as an analogy; you’re not going to get any stronger by using a forklift all day, but you’re also not going to keep up with other operators if you’re loading trucks by hand. Strength can also make you a better forklift operator (tighter ratchet straps, convenient staging, etc.) but using a forklift for everything all the time will make your body weaker.
I think it’s analogous, and I think if you’re not using your brain as much as you would without AI, you’re not using it as a tool and are instead outsourcing your thinking.
Nobody expects a mathematician to do arithmetic by hand, but also nobody will beat a good mathematician without a calculator at arithmetic, either. I see it as generally the same thing; part of the process is (relatively) automated allowing for higher level development.