4 ms·

Let’s look at what people are actually arguing: that chatbots might be sentient because their writing reads like the writing of a sentient human. That’s it. So

Let’s look at what people are actually arguing: that chatbots might be sentient because their writing reads like the writing of a sentient human. That’s it. So since the criteria being put forth is similarity to humanity (in the form of writing), we should also consider all the ways in which chatbots are not similar to humanity (in the form of almost everything else). We don’t need a general theory of sentience. And I’m not arguing that sentience in general has to look like a human. We just need to understand the narrow scope of the current arguments for chatbot sentience.

corobo 4y ago
> because their writing reads like the writing of a sentient human Does it even do that? Is there anyone else out there that can tell if it's AI within the first two sentences or so? e.g. https://news.ycombinator.com/item?id=34979687 https://news.ycombinator.com/item?id=34979687 It feels super obvious when it's an AI response.. I can't quantify it, it's just.. obvious. There's some patterns or something in there I'm picking up on, and whatever pattern matching I'm doing with my electric meat seems to be recreate-able using AI itself: https://openai.com/blog/new-ai-classifier-for-indicating-ai-written-text https://openai.com/blog/new-ai-classifier-for-indicating-ai-...