4 ms·

Does this increase anyone else's sense that LLMs are actually modeling something similar to the way humans think? It certainly does for me. I find that negatio

by akavi 3y ago

Does this increase anyone else's sense that LLMs are actually modeling something similar to the way humans think? It certainly does for me. I find that negations/conjugations of negations are surprisingly difficult for to parse. Beyond two negations at most, I can feel myself having to switch into "code" mode, where I'm explicitly casing out the booleans, to understand them. Obviously the examples given in the article are simpler than double negations, but it feels very consistent with the idea that "negations are an inherently hard part of language processing".

jprete 3y ago
This specific example doesn’t, but I think there’s a fair amount of brain processing similar to what LLMs do. I just don’t think it’s the part of our mental experience that we regard as making decisions; it’s more like the part that is active when absent-mindedly daydreaming.
leptons 3y ago
>are actually modeling something similar to the way humans think? Apparently humans can't think enough to write a paragraph anymore, so they have to ask ChatGPT to do it for them. At what point does ChatGPT ask ChatGPT to write it a paragraph? At that point maybe LLMs are actually thinking like humans.
furyofantares 3y ago
"Don't think of a pink elephant" comes to mind with this stuff as well, where telling someone not to think about something causes them to think about it.
- tgv 3y ago
  That's different. That's a meta-linguistic joke, that works because we can't read that sentence without at least briefly thinking of a pink elephant (for some value of "thinking" anyway).
canjobear 3y ago
Here's a good example of the difficulty of processing negation for humans. "No head injury is too trivial to ignore" -- What does this mean?
mirekrusin 3y ago
Yes, subconscious/emotions don't have concept of negation, they do have fear, projections of pain etc. instead.
bluepoint 3y ago
The negations have multiple roles. An important one is that they add “color saturation” to the words. “It is bad”, “It is good”, “It is not bad”, “I can’t say it’s good”, “I can’t say it’s not bad”, are all different statements. So negations are difficult, maybe even for humans…

by akavi 3y ago

jprete 3y ago

This specific example doesn’t, but I think there’s a fair amount of brain processing similar to what LLMs do. I just don’t think it’s the part of our mental experience that we regard as making decisions; it’s more like the part that is active when absent-mindedly daydreaming.

leptons 3y ago

>are actually modeling something similar to the way humans think? Apparently humans can't think enough to write a paragraph anymore, so they have to ask ChatGPT to do it for them. At what point does ChatGPT ask ChatGPT to write it a paragraph? At that point maybe LLMs are actually thinking like humans.

furyofantares 3y ago

"Don't think of a pink elephant" comes to mind with this stuff as well, where telling someone not to think about something causes them to think about it.

tgv 3y ago
That's different. That's a meta-linguistic joke, that works because we can't read that sentence without at least briefly thinking of a pink elephant (for some value of "thinking" anyway).

canjobear 3y ago

Here's a good example of the difficulty of processing negation for humans. "No head injury is too trivial to ignore" -- What does this mean?

mirekrusin 3y ago

Yes, subconscious/emotions don't have concept of negation, they do have fear, projections of pain etc. instead.

bluepoint 3y ago

The negations have multiple roles. An important one is that they add “color saturation” to the words. “It is bad”, “It is good”, “It is not bad”, “I can’t say it’s good”, “I can’t say it’s not bad”, are all different statements. So negations are difficult, maybe even for humans…