Weak comparisons help no-one, photoshop is nothing like LLM’s
All of the big commercial LLM’s (without exception afaik) have been trained on a large corpus of data that has been obtained by various sketchy and illegitimate means. (some legitimate as well).
That’s the major difference between the two.
If you are using a model that has only been trained on legally obtained data, disregard this point.
I’m not even against competent tool use of LLM’s but please use better arguments.











TL; DR;
I agree, how does throwing out bad comparisons relate to that ?
if you are talking about cascadeur or something similar, that doesn’t use an LLM afaict, it’s based on ML Trained on their own internal data (or so they say).
I don’t disagree that tools used in a way that plays to their strength are useful.
People are often conflating AI with LLM’s, which makes sense for the average person, because that’s how it’s been marketed and sold.
LLM’s aren’t even really AI but here we are.
I was very specific in my wording, but as i said, i could be wrong, if you can point to any big commercial LLM’s that don’t adhere to my classification i will concede the point.
I mean, yes, that’s what i said.
So i stand my my conclusion that in the context you laid out, Photoshop isn’t a good comparison to most, if not all of the current tools that would be considered AI.
I could be wrong but half of that statement was sarcasm.
I basically read it as:
I think the fundamental misunderstanding here is how the term AI is used.
None of these things are really intelligent and LLM’s are predictive semi-hallucination machines cobbling together best guesses at what’s supposed to come next in the sequence.
The way i personally see it is that the latest gen “AI” stuff is basically sitting on LLM’s in some capacity. Area recognition, language, image/code generation etc.
Anything else is just normal(perhaps smart) tools, using algorithms of some kind, ML etc