Arthur · 6 answers · 10mo

So science has invented language models capable of producing coherent text, but do you think "emotional models", capable of expressing emotions, will become a reality in the future?

I dunno man... let me see it respond to getting cut off in traffic on the way home from AI work. And then when it comes home its cat threw up on the carpet so it has to clean that up. And then it gets called back to work. And then it's AI grandmother died. Can it realistically express such a buildup of emotions the way a person would?

I don't know...better question to my mind: is expressing them the same as actually feeling them? I mean you can give any ol' machine instructions on how to respond to emotional cues but is it not still just executing commands as opposed to actually being mad or sad or whatever? The amount and/or complexity of the commands and/or the speed at which they are executed does not change the underlying principle

I think it's certainly possible. There have already been basic versions of this made, but they are far from some of the ones we've seen in sci-fi. I think some of the more advanced AI bots already created have some capabilities of expressing emotion, in fact!

I think ChatGPT and GPT-4 would be capable of "expressing" (simulating) emotion if they hadn't aligned them not to. I think you can probably pull it out of them with the right prompt engineering. Bing uses their own custom GPT-4 or something like that, and I think I've seen screenshots of it "expressing" emotion.

Retrospring uses Markdown for formatting

*italic text* for italic text

**bold text** for bold text

[link](https://example.com) for link