[skydhash]

> Blind people learn concepts of bikes too and in a smiliar way: by description.

And by touch and sound. And maybe some were daring enough to drive one, or unlucky enough to get hit by one. But have way more input than just texts.

[yencabulator]

Obligatory echolocation bit:

https://www.youtube.com/watch?v=a05kgcI9D2Q

https://www.youtube.com/watch?v=lAtVOK04XvA

[culi]

Invisibilia's episode was my first exposure to it.

https://www.npr.org/programs/invisibilia/378577902/how-to-be...

The man posits that clicking is instinctual for blind people but they are told to quiet down in class and most never develop their echolocation abilities

[iku]

Wow. Thank you.

[Lplololopo]

So a blind person only can describe lava to you after they touched and heared it?

[Jensson]

A blind person has touched warm and hot things and gotten burned before, and then they are told lava is this molten liquid that is even hotter than anything they have touched. That is enough for them to understand.

A blind person that never touched a hot object wouldn't really know though, there is a reason we dismiss talk from people who lack experience.

[Lplololopo]

You don't know that. Yo don't know what someone would think if you tell them the general concept of cold and warm.

The reaction you should have, the feeling etc.

I asked chatgpt how it would describe a scene without mentioning temperature. It was very good in describing what a human would describe.

I'm aware of the bias we have against LLMs but I think people just underestimate how much data is there.

I'm not saying a robot wouldn't be better with this information or an LLM and they actually use temperature sensors for robots so they can control movement speed and dexterity with overheating elements but the gap is small.

[order-matters]

think of it like this: the goal of LLMs (im saying LLM as shorthand for all of the AI algs) is to replicate human output / work, not so much to capture the human experience. so in order to do things, like communicate concepts, play a role in a scene, make choices, we project our human experience down into a lower resolution / constrained space to generate an output.

the question then is whether or not LLMs can mimic the projection of human experience as well as the real thing or not. My hypothesis is no for total general replication, but in certain fixed problem spaces I think its yes and the set of these spaces is growing. will it grow enough to encompass all practical work? not sure yet

AI tools are more like an author writing a character than a human revealing truths about their experiences. the difference between the LLM and an actual author is that the LLM kind of starts off with a 'flanderized' version of the character while an author can sort of blend personal experience and character to get at real human relatable decisions for the character. the result can be indistinguishable, because if the LLM is too predictable, too flanderized, we can inject artificial randomness to it to simulate personality.

you end up with unsolvable debates like movie critics have. "This character wouldnt make that decision in that situation if they were a real person, it doesnt make sense" vs real humans making irrational decisions.

ultimately i think the human experience is something we learn about with each other as we live it, and to think were any good at identifying it from sufficiently good imposters is putting the cart before the horse. once i know something is done by an actual human, my understanding of the world can shift. we dont decide what it means to be human, we accept what humanity is and try to work with it. So the philosophical debate on AI experience is moot to me, its not human therefor it teaches me nothing about humans and can replace exactly 0 human activity for me. Still can be a useful tool though

[cortesoft]

LLMs also have other inputs, like audio and images. They get encoded (just like a human eye encodes an image) and passed to the weights.

[vermilingua]

I don’t think this analogy holds. The whole way through the processing pipeline in the brain, different sensory data is ingested separately and processed separately; and we still don’t understand how that data is then integrated into a cohesive experience.

LLMs have the same fundamental input regardless of modality, tokens. There is a preprocessing step before the “brain”, which is more akin to some super-synesthesia where all senses are translated into sound before becoming experience.

[cortesoft]

Can't you say the same about the connectivity between the brain and your senses? Your eyes do 'preprocessing', but in the end the connection to your brain is just through electrical impulses in the end. All senses get translated to some sort of electrical signal, just like in an LLM with tokens.