|
|
|
|
|
|
by trifurcate
1222 days ago
|
|
|
The word you are looking for is an _embedding_. Embeddings are to language models as internal, too-rich-to-be-fully-described conceptions of ideas are to human brains. That's how language models can translate text: they have internal models of understanding that are not tied down to languages or even specific verbiage within a language. Probably similar activations are happening between two language models who are explaining what a "promise" means in two different languages, or two language models who are telling different stories about keeping your promise. This is pattern recognition to the same extent human memory and schemas are pattern recognition, IMO. Edit: And for the rest of your post: > Which AI program will make a promise to you? When it fails to fulfill its promise, will it feel bad? Will it feel good when it keeps its promise? Will it de-prioritize non-obligations for the sake of keeping its promise? Will it learn that it can only break its promises so many times before humans will no longer trust it when it makes a new promise? All of these questions are just as valid posed against humans. Our intra-species variance is so high with regards to these questions (whether an individual feels remorse, acts on it, acts irrationally, etc.), that I can't glean a meaningful argument to be made about AI here. I guess one thing I want to tack on here is that the above comparison (intra-species variance/human traits vs. AI traits) is so oft forgotten about, that statements like "ChatGPT is often confident but incorrect" are passed off as meaningfully demonstrating some sort of deficiency on behalf of the AI. AI is just a mirror. Humans lie, humans are incorrect, humans break promises, but when AI does these things, it's indicted for acting humanlike. |
|
|
I would phrase that same statement slightly differently:
"they have internal [collections of activation weightings] that are not tied down to languages or even specific verbiage within a language"
The phrase "models of understanding" seems to anthropomorphize the ANN. I think this is a popular way of seeing it because it's also popular to think of human beings as being a collection of neurons with various activation weightings. I think that's a gross oversimplification of humans, and I don't know that we have empirical, long-standing science to say otherwise.
> This is pattern recognition to the same extent human memory and schemas are pattern recognition, IMO.
Maybe? Even if the embedding and the "learned features" in an ANN perfectly matched your human expectations, I still think there's a metaphysical difference between what's happening. I don't think we'll ever assign moral culpability to an ANN the way we will a human. And to the extent we do arm ChatGPT with the ability to harm people, we will always hold the humans who did the arming as responsible for the damage done by ChatGPT.
> All of these questions are just as valid posed against humans. Our intra-species variance is so high with regards to these questions (whether an individual feels remorse, acts on it, acts irrationally, etc.), that I can't glean a meaningful argument to be made about AI here.
The intra-species variance on "promise" is much, much lower in the mean/median. You may find extremes on either end of "how important is it to keep your promise?" but there will be wide agreement on what it means to do so, and I contend that even the extremes aren't that far apart.
> Humans lie, humans are incorrect, humans break promises, but when AI does these things, it's indicted for acting humanlike.
You don't think a human who tried to gaslight you that the year is currently 2022 would be indicted in the same way that the article is indicting ChatGPT?
The reason the discussion is even happening is because there's a huge swath of people who are trying to pretend that ChatGPT is acting like a human. If so, it's either acting like a human with brain damage, or it's acting like a malevolent human. In the former case we should ignore it, in the latter case we should lock it up.