Most AI outputs are useless and annoying

[zurew]

Similar issues come up with trying to solve RR with predefined rules.

Every rule requires a specification in its application and every rule requires an interpretation. The number of ways you can interpret and apply any given rule is combinatorially explosive and context sensitive.
A rule always conveys more than it explicitly says (You cannot specify all the  condition of the application of the rule in the rule).

"Okay lets try to make a rule for how to use this specific rule" - You are just pushing back the issue one more step, because now you have the exact same problem (you need to know all the specificity conditions how to apply that higher order rule), so your ability to follow rules is based on something other than rules.

 

 

RR also goes deeper than representations (mental entity that stands for or refers towards an object in the world) , because representations doesnt explain RR, they presuppose it.  

Empirical evidence for this: "Things and Place - How the mind connects the world by Zenon W. Pylyshyn" - The multiple object tracking, where the more objects you need to track at the same time, the more you lose your ability to track the attributes of said objects - So for example, lets say there are 8 objects and one of them is a red X. After lots of momevent, you can still tell the location of that object, but you won't notice that its content has changed over time (the red X has become a blue circle). The only thing that was consistently tracked of that object is its here-ness and now-ness (or in other words adverbial qualia). So all the representational aspects of the object was changed, but the ability to track it still remained.

 

Edited 19 hours ago by zurew

[Girzo]

1 hour ago, AerisVahnEphelia said:

You asked a wolf one, that's what I did, I could have got the monkey one. ( I mean you asked more of a wolf style, but I just did a tail, but that's probably possible to design him in a wolf way )

No, I have said the AI is reterded by trying to make Goku into a dog when I DON'T want to, just because there's dog in the pose's name. That was the situation with Midjourney last year or whenever I was seriously trying it out. It just kept mixing in unrelated meanings from text prompts, or changing poses in so many ways.

It was impossible for me when I tried for him to have no wolf characteristics, and when using the image prompt with the pose it always changed the pose so it was useless, but your first try is what I have always wanted and were unable to procure.

[Emerald]

AI is useful within a limited context. But I feel like people project the expectation onto LLMs that it's ACTUAL artificial intelligence.

But it isn't artificial intelligence at all. We are a LONG way off from discovering how to actually create real AI... if it's even possible.

LLMs only have two skills among thousands that the human mind possesses.

One of those skills is data collection and retention. The other skill is holistic data synthesis to create text, sounds, and images.

And because these are things that human minds can do... sometimes better than LLMs and sometimes worse than LLMs... it gives the illusion that we're interfacing with an artificially created intelligence when we're interfacing with LLM software.

But it's really just another computer program like any other. Only this one has been trained to collect data and holistically synthesize it into new text, sounds, and images.

And the text, sounds, and images it creates resembles the texts, sounds, and images that humans create... sometimes identically or just in relation to general aesthetics and style.

So, it gives the illusion of intelligence and sentience. But it's just an illusion.

So, it's only annoying if you don't expect too much from it. 

[Emerald]

On 10/13/2025 at 8:55 AM, ryoko said:

It's just a matter of time before we have truly self learning AI, all with unique experiences.

LLMs simply cannot do that. And that kind of program can NEVER be developed enough to learn to do that, as it isn't how LLMs work.

If we take for granted that it can be done... we'd need to invent something TOTALLY different than LLMs.

[Hello from Russia]

On 10/12/2025 at 2:16 PM, Leo Gura said:

It is more intelligent than most humans simply because it has unlimited memory for dry factual information. This creates an illusion of intelligence in the context of saying factually known things, like historical facts or technical scientfic knowledge.

But it cannot think truly novel ideas. Only recombitions of existing ideas.

It can quote science to you, but it can't do new science. It can quote philosophy to you but it can't actually do philosophical work.

99% of humans probably can`t do new science too (:

You are right about it being a sort of encyclopedia

But also it's pretty unbiased and fairly open minded / non dogmatic. I enjoy to have conversations much more with AI than with most humans for this reason

Sure, its not perfect and has some subtle "sciency" built-in bias, but still compared to the average person it's just night and day

It can even cheer you up or console you much better than most humans and you even perceive it in a much more authentic way because it doesnt really has any hidden agenda towards you (though who really knows) 

[ryoko]

3 hours ago, Emerald said:

we'd need to invent something TOTALLY different than LLMs

Yes. I'm using the term AI and LLMs very mindfully. AI implies future possibilities like embodied AI, not LLMs.

Edited 1 hour ago by ryoko

[Carl-Richard]

4 hours ago, Emerald said:

And because these are things that human minds can do... sometimes better than LLMs and sometimes worse than LLMs... it gives the illusion that we're interfacing with an artificially created intelligence when we're interfacing with LLM software.

Yeah, it's like it produces text kinda like a human, therefore it must be conscious like a human (is what absolutely braindead materialists like Mike Israetel think). Meanwhile you don't hold a dog or a human baby to the same standard. It's such a piecemeal and childishly short-sighted understanding of consciousness.