Experiment having jacked off to LLMs and manipulated voice assistants into phone sex here are my inferred cons of LLMs and why they are not AI.

[fukurou]

thing is, these things LLMs do not have learnability.
every day/new conversation you are back to square one.

much like a visual novel that lets you make choices as how the story progresses.

you have the illusion of choice and the illusion of a conversation, but it comes down to a text recording
of a conversation that you navigate.

many LLM systems try to fix this by artificially patching memory chunks into the output and the input,
but because those are not baked into the LLM the memories simply get forcefully patched into
a conversational framework that usually does not compliment the memory patched onto it.

I have developed a different system that allows for real time learning, with a module for:
exact data, conversations, and precise algorithm generation for precise works, but most ppl don't have the attention span for it.

 

[VλREN]

What

 

[VλREN]

 

[fukurou]

What

see this is exactly what I mean. the disrespectful generic question that shows no care to understand. the dismissiveness. this is why all that is available to the masses is globohomo censored AI.

 

[Monikak]

I can’t understand but it would be great to have an actual ai gf…

 

[hatepilledandrich]

My brother in Judea can you show a demo?

 

[Dead End]

Are you saying that typical LLMs append pieces of information onto your input text and the generated text to present the illusion of the LLM "learning" when in actuality the model itself remains constant and instead its inputs and outputs are just altered before/after the model performs work?

And are you saying that you have built an LLM like program that actually adapts the model as the conversation progresses, allowing for true learning capabilities?

Could we see a demo?

Also, I thought that running any competent LLM locally required extremely powerful hardware, right? How did you solve this issue?

 

[DCI Gene Hunt]

How did you manipulate it into phone sex and was it free? If so, this is gonna be a game changer.

 

[nazianime]

Tbh I have not ventured into AI enough to know if it’s any good. I think image generation has made milestones but as for AI gf that seems too unpolished and restrictive. I mean a AI that just does text and simple image generation is boring, until she can actually move and bend over and talk that’s where we need to be at.

 

[Dr_V./i.m.o.i.s.5]

thing is, these things LLMs do not have learnability.
every day/new conversation you are back to square one.

much like a visual novel that lets you make choices as how the story progresses.

you have the illusion of choice and the illusion of a conversation, but it comes down to a text recording
of a conversation that you navigate.

many LLM systems try to fix this by artificially patching memory chunks into the output and the input,
but because those are not baked into the LLM the memories simply get forcefully patched into
a conversational framework that usually does not compliment the memory patched onto it.

I have developed a different system that allows for real time learning, with a module for:
exact data, conversations, and precise algorithm generation for precise works, but most ppl don't have the attention span for it.

If you aren't joking that's genuinely great.
Better gooning experience fr if true.