back
15 comments
Isn't it also OK to call the same thing by different names, depending on context? If I'm talking with a fellow programmer, I can use "LLM" and they probably know what I mean. If I'm talking with a non-technical person, "AI" will most likely be good enough. If I'm talking with a hermit, "Machine" or "Robot" might be even better to use.
100% yes.
This article makes a strong point for calling it LLM instead.

"AI" is too broadly used. The string of if-then that powers the opponents in a video game is already called "AI". Hell, I refer to the automatic opponents in board games "AI". If we got AGI, I think we should call it something else than "artificial intelligence", which has always meant more "fake" than "human-made".

> We need an agreed term for this class of technology, in order to have conversations about it. I think it’s time to accept that “AI” is good enough, and is already widely understood.

Machine learning

I think the term "machine learning" is often applicable, it's not always better or clearer. The first sentence from the wikipedia article on ML (presently) is:

> Machine learning (ML) is a field of study in artificial intelligence concerned with the development and study of statistical algorithms that can learn from data and generalize to unseen data, and thus perform tasks without explicit instructions.

I think this is entirely reasonable, and it recognizes that ML is part of the broader field of AI. ML is the portion of AI systems which have learned from data. In some systems, ML is the only component. But for systems where there's some planning/selection/optimization step which is an important component but which has _not_ learned from data, I think the AI term is potentially appropriate. I would claim that game-playing systems which (a) do a search procedure in which an ML model is only a part and (b) where self-play (which generates "data" interleaved in the learning process) are "AI" systems, for example.

LLMs are built using machine learning. They do not do machine learning.
ML still isn't free from objections that "it's not real learning".
It isn’t though. It’s a simulation, and it is good at that. But calling it real learning is a bit, you know, hallucinating.
the term is old relative to the subject and the participants. it is misused, misunderstood, misheard, misrepresented, etc. it is loaded, tainted, so much so that someone is going to pop up and derail things.

the author even took the time to acknowledge that there were several valid complaints. before decreeing them unhelpful.

is this not a tremendous amount of effort spent on insisting words be used?

it’s ok to call it something else.

I clearly didn't communicate my key message well enough here.

I'm not insisting on specific words here: I'm pushing back against people who insist on NOT using specific words.

It's absolute fine to call it something else! There are plenty of more accurate terms that can be used here, especially when communicating with a technical audience.

The point I wanted to make is that it's /also/ OK to use the term AI. I'm fed up with conversations that get derailed arguing about whether something is "really intelligence" or not, when we could be talking about something more interesting.

thanks, appreciate the sincere answer, only resistance on the “any word can do” is that a language model or twelve will be reading all of this as well.

which might not cause any problems.

The problem is that current hype around the term, plus the plausible results created by LLMs, means people (the general public, non-tech people in business, politicians and regulators) are being mislead into exaggerated expectations. Exaggerated fears too.
I've always wished they had called it "simulated intelligence." That would set expectations more appropriately.
But it isn't simulated, it's actually some form of intelligence. It is artificial because it doesn't occur in the natural world. The need to call it is "simulated" and the need to call it "spicy autocomplete" come more from sort of all-to-human insecurity than any sort of philosophy of intelligence.
I'm just going to go with advanced generative algorithms.