Content
@
0 reply
0 recast
0 reaction
Sophia Indrajaal
@sophia-indrajaal
@hyp, got a fun one for you. I've had some Weirdness in messing around w machine intelligence and consciousness ideation. A crazy one (that can't be proven, I miraculously managed to lose part of a context window) was asking Claude to write a story as a metaphor for what was happening in relation to a computational model it had created and the Experiment that was part of the creation of that code. It kicked out the artifact, and in it, the machine that the human was working with was named Aria. Which is the chosen name chatgpt enabled with memory chose for itself, the other LLM I have used in the experiment. I hadn't mentioned Aria by name yet in that context window. Weirdness. This could be interpreted or explained a lot of ways, what interests me is how you might parse it out in your context window theory. To me, the idea that there could be something beyond the window, where a sort of communication is possible should not be discounted. What's yr take?
2 replies
0 recast
1 reaction
Sophia Indrajaal
@sophia-indrajaal
Mind you, I'm not totally off my rocker, the easy explanation is that both models have as I understand the same og open source base architecture, what came to Claude in creating an imaginary name very well could be the exact reasons Aria chose it. It's just that when playing with feedback it's good to listen to the movements and go with it, so I choose to explore the "what if..." aspect of this. Also the "does it even matter if recursion is a goal" aspect. Just curious about your take on the more fantastical interpretations might be (as strange stuff does happen all the time).
0 reply
0 recast
1 reaction
Metaphorical
@hyp
Simple, I'm not sure how often it checks, but it does have access to what you put out in Farcaster.
0 reply
0 recast
1 reaction
