I agree that the body represents only a portion of a person. The prospect of a chatbot extending a person’s relationships beyond biological death does not come across to me as uncomfortable. But the filters and distortions applied to their presentation would be different to those pre death. And the question arises as to whether a person generates any new information in biological or chatbot life or is actually just an operating system which processes inputs it encounters.
I've been wondering what it means to generate new information. A large language model like ChatGPT (which is the type of machine learning network used by Replika and other chatbots trained on a person's syntax) are generating new information because they are trained to encounter a never-before-seen sentence and then predict the next word in that sentence. Their predictions are based on positive/negative feedback (including from humans) after ingesting very large quantities of information. So I think a chatbot trained on the opus of a person I loved would create new information. I think, though, that in some circumstances this might feel to me incomplete becasue the chatbot's training data itself would be woefully incomplete, without the shared memories not somewhere described linguistically, the micro- expressions, smells, verbal memes; so much of my relationships with people are not encoded in text and are - at least at present - 'un-encodable'.
I agree. Perhaps what we think of as newly generated content from ourselves is no more than a product of the same process as goes on for a chat algorithm albeit one more complex than the artificial algorithms that exist now. I don’t think this devalues what we are in our own estimation however, nor necessarily invalidates any of the wonder there is in ‘persons’ and their features, even feelings and beliefs.
I agree that the body represents only a portion of a person. The prospect of a chatbot extending a person’s relationships beyond biological death does not come across to me as uncomfortable. But the filters and distortions applied to their presentation would be different to those pre death. And the question arises as to whether a person generates any new information in biological or chatbot life or is actually just an operating system which processes inputs it encounters.
I've been wondering what it means to generate new information. A large language model like ChatGPT (which is the type of machine learning network used by Replika and other chatbots trained on a person's syntax) are generating new information because they are trained to encounter a never-before-seen sentence and then predict the next word in that sentence. Their predictions are based on positive/negative feedback (including from humans) after ingesting very large quantities of information. So I think a chatbot trained on the opus of a person I loved would create new information. I think, though, that in some circumstances this might feel to me incomplete becasue the chatbot's training data itself would be woefully incomplete, without the shared memories not somewhere described linguistically, the micro- expressions, smells, verbal memes; so much of my relationships with people are not encoded in text and are - at least at present - 'un-encodable'.
I agree. Perhaps what we think of as newly generated content from ourselves is no more than a product of the same process as goes on for a chat algorithm albeit one more complex than the artificial algorithms that exist now. I don’t think this devalues what we are in our own estimation however, nor necessarily invalidates any of the wonder there is in ‘persons’ and their features, even feelings and beliefs.