Further to my dislike of the mirror test, I have another observation.
I use LLMs to air my thoughts and often get useful expansion and feedback. But in the deeper moments I have to remember that this is a machine, a language-bound entity built entirely on digesting human written interactions, pure pattern-matching from a framework. It has no outside, no life to give or receive context to those words.
At this point I reminded myself – when I bounce ideas off it, there is no man in the mirror. The LLM as mirror itself is just as much an illusion as mirrors are illusions.
So when even Richard Dawkins can report results which could be interpreted as him ‘failing the mirror test’ in this regard, should we downgrade humanity’s sentience level?
Leave a Reply