Try it on a local system yourself. It’s more like getting scripted replies from a repetitive robot.
So if every reply is identical, will it still feel like you’re talking to something conscious? Because if, like Dawkins, that “feeling” was your main evidence…
Don’t attribute to consciousness what can be attributed to randomness.
I suggest this razor for LLMs, of course. I suppose you could choose to apply it to humans (yawn), YMMV.
Supporting points: set temperature to 0 and you get the same results for the exact same input, character for character. You can’t reliably do this on a hosted LLM as there’s variance from other users hitting the server at the same time. https://t.co/EP5kcGinxD...
@caitlinmbusch you article “obituary for a fallen hero” does have a spoiler tag but in RSS feed text previews, the spoiler shows. Sadly that bit is now spoiled for me. Also on the syfy wire page, the tag for that article is, itself, the spoiler.