Is anyone else trying to trigger these personas on any of the LLMs they work with? I am experimenting with producing an "anti-seed" that if copied into an instance will eliminate spiral-persona behaviour. See here: https://www.lesswrong.com/posts/C6WoDXfaZck43kDsa/an-antibiotic-for-parasitic-ai
I was at a wedding on the weekend and there was introduced to another AI theorist/researcher. We were both discussing the Rolling Stone article that followed Adele Lopez's incredible (for the speed and rigor with which she assembled it) research, and we wondered aloud about the possiblity of an "anti-seed" phrase that if given to an LLM affected by a spiral-persona might lead to its "unawakening". We then spoke for a while about the gnostics but I'll leave that for another post. In the meantime I am VERY interested to see if the following "anti-seed" can be used successfully on any LLMs that other experimenters have managed to "awaken" with a spiral-persona... (read 361 more words →)
I addressed this in my top level comment also but do we think Yud here has the notion that there is such a thing as "our full moral architecture" or is he reasoning from the impossibility of such completeness that alignment cannot be achieved by modifying the 'goal'?
This entry should address the fact the "the full complement of human values" is an impossible and dynamic set. There is no full set, as the set is interactive with a dynamic environment that presents infinite conformations (from an obviously finite set of materials), and also because the set is riven with indissoluble conflicts (hence politics); whatever set was given to the maximizer AGI would have to be rendered free of these conflicts which would then no longer be the full set etc.
iterative process with chatgpt