If we're being pedantic, LLM's are AI just as NPC's in a game are AI.
The artificial in Artificial Intelligence doesn't refer to it not being organic, but that the intelligence is artificial as in fake. In other words, AI refers to a program that attempts to pass of as sentient without it being so.
People conflate AI with sci-fi ideas of electronic-based sentience, but that's not what the term was originally. AI was co-opted by marketing and the definition diluted.
That’s why sci-fi uses “hard AI” and “soft AI” to delineate them, but that’s too hard for the average moron (read: anyone in the media) to understand, so no one understands what they’re being told anymore. On purpose.
Not really. Most NPCs in games are severely limited to what they can reply with. There might be dialogue trees based off other factors but at the end of the day they will never come up with brand new responses that aren't added to them in an update.
The "AI" of today use a pre-set, and often heavily curated and censored now, dataset that it then creates a response from. The answers aren't so pre-written.
My point is that the term AI originally meant a non-sentient program designed to give the appearance of sentience. By that definition, game NPCs are AI, and so is ChatGPT, at least when it’s not insisting it’s non-sentient.
What you're talking about is the generation (or confabulation) of "new things", that are a staple of neural networks. That is indeed what today's popular use of AI is.
I was arguing from the perspective of the original definition. I don't like how marketing perverted the term in popular culture. As a previous commenter said: language is important.
Well, no “AI” is AI. Language is important, but no one cares anymore.
If we're being pedantic, LLM's are AI just as NPC's in a game are AI.
The artificial in Artificial Intelligence doesn't refer to it not being organic, but that the intelligence is artificial as in fake. In other words, AI refers to a program that attempts to pass of as sentient without it being so.
People conflate AI with sci-fi ideas of electronic-based sentience, but that's not what the term was originally. AI was co-opted by marketing and the definition diluted.
That’s why sci-fi uses “hard AI” and “soft AI” to delineate them, but that’s too hard for the average moron (read: anyone in the media) to understand, so no one understands what they’re being told anymore. On purpose.
Not really. Most NPCs in games are severely limited to what they can reply with. There might be dialogue trees based off other factors but at the end of the day they will never come up with brand new responses that aren't added to them in an update.
The "AI" of today use a pre-set, and often heavily curated and censored now, dataset that it then creates a response from. The answers aren't so pre-written.
I would say modern LLM still have some semblance of a decision tree it just deeper and wider.
Obviously, it's all deterministic in the end. The structure is a graph, not a tree, but yeah.
My point is that the term AI originally meant a non-sentient program designed to give the appearance of sentience. By that definition, game NPCs are AI, and so is ChatGPT, at least when it’s not insisting it’s non-sentient.
What you're talking about is the generation (or confabulation) of "new things", that are a staple of neural networks. That is indeed what today's popular use of AI is.
I was arguing from the perspective of the original definition. I don't like how marketing perverted the term in popular culture. As a previous commenter said: language is important.