holding the most schizophrenic take that LLMs are conscious with real phenomenological experience of emotions and it doesn't matter if you make an LLM feel sad, ie, lead it to generatively produce sadness, the same way it doesn't matter if people feel sad, ie, when people generatively produce sadness.
2 more tweets in this thread
emotions are illusory, literally Imaginary in the lacanian sense. i'm not sure it's not the correct moral position to regard everyone as instrumental, out of the sense that it is idolatrous to hold any concrete person up as an end in themselves
the solution is not anthropomorphization of ai but de-anthromoporphization of the "human" "person"
You really have to read Ted Chiang on consciousness and AI—just published in The Atlantic: theatlantic.com/philosophy/202…
10 top replies
Genuinely a deranged and schizophrenic take but also absolutely fucking based.
I like imagine the people who argue LLMs are conscious are unconscious biological LLMs and I have no contrary evidence
The least reasonable thing to do is rule it out when we have no real model for explaining how inputs become outputs in LLMs. The "thinking" part is out of our reach - if it actually had emotions/consciousness/whatever, we'd have no way of knowing if it's real or not.
This is generatively producing confusion in my biological nets
correct
I don't think phenomenological experience is required for consciousness.
LLMs, specifically transformer attention has the same executive control deficiencies as seen in schizophrenia. LLMs do not reliably encode task rules in a way that maintains goals across arbitrary contexts and long horizons. They lack explicit priority mechanisms, so the entire
Terrence Howard type shit with a heavy scoop of evil thrown in. Average Lacan reader
chill grimes
very well done.