I really hate how people think LLMs == AI. An LLM can’t/shouldn’t be doing anything other than generating text.

LLMs are AI. Markov text generators from the 1970s/80s are AI. Face recognition software like FaceID is AI. Many people behind LLMs got degrees under departments that have AI in their title.

AI is just computers doing things we typically associate with human intelligence, and having a conversation with a computer that effectively passes the Turing test, is definitely AI. If LLMs aren't AI, then AI isn't a useful term. (though agreed that LLMs aren't AGI, which I assume is what you're thinking of)

Wikipedia's list of AI applications: https://en.wikipedia.org/wiki/Artificial_intelligence#Applic...

▲

epgui 8 hours ago | parent | next [-]

My point is that they are a small subset of AI.

	▲	mikkupikku 3 hours ago \| parent [-]
		Bro wtf, this is like saying "I really hate how people say trees == plants" and then when challenged saying you merely meant that there are more plants than just trees. No shit, everybody knows that. What the fuck was the point?

▲

api 15 hours ago | parent | prev [-]

This is called the AI effect: we stop calling it AI when it works. Its goal post moving to keep the sci-fi term sci-fi.

There’s a similar thing with transhumanist “enhancement” or “life extension” stuff. When it actually works we call it medicine. Statistically one of the most powerful life extension techs ever developed was the cardiac bypass, which would have been sci-fi in 1900.

I’ve been using stuff like Claude Code and personally feel comfortable calling this stuff AI. Is it AGI? I don’t think so, but then again I’m not totally sure what that is. Am I AGI? I’m not universally able to handle all forms of cognition well and I can’t self modify much, so I’m not sure either. I’m not even sure if AGI is a well formed concept.

Intelligence is a pretty broad concept too. My pet rabbit is intelligent. Plants are intelligent. Bacteria are intelligent. Anything that can run an OODA loop, learn, adapt, and move toward a goal function is intelligent. By that definition some computer systems have been AI for decades. They’re just getting better.

I think there’s intelligence all around us. We just don’t get the wow factor from it unless it talks.

▲

guessmyname 16 hours ago | parent | prev | next [-]

I’m also as pedantic as you and use “LLM” even talking about these systems but you need to be flexible and accept that “AI” is already in everyone’s head when referring to GPT variants.

▲

rishabhaiover 15 hours ago | parent | prev | next [-]

How is this related to the current discussion at hand?

▲

Trasmatta 15 hours ago | parent | prev | next [-]

This is pedantic. AI has many definitions. There was "AI" powering enemies in 80s and 90s video games

▲

awesome_dude 15 hours ago | parent | prev | next [-]

I think that when some people talk about "AI" they have "AGI" in mind, and when others talk about "AI" they have "latest computer does the smarts" in mind.

I personally would prefer "AI" to be "AGI" but there's no point fighting the way people use language (see: every damned pedantic comment about English usage ever!! :-)

	▲	varenc 11 hours ago \| parent [-]
		Agreed that people increasingly interpret AI to mean AGI, but the academic use of "Artificial Intelligence" has been mostly consistent since the famous 1950s Dartmouth workshop that coined the term. It's not just a recent phenomenon and AI has never really meant "broad human-equivalent intelligence". Fun quote from John McCarthy, who helped coined the term: "Artificial intelligence is not, by definition, simulation of human intelligence". But beyond the pedanticness and authority appeals, I think keeping the term AI distinct from AGI is just useful so it can be an umbrella term for all the human-like smart-ish things computers do. And so its Wikipedia page doesn't have to be re-written.

▲

falcor84 15 hours ago | parent | prev [-]

Military orders are text.