OUR PROGRESS WITH ALIGNING AI WITH TRUTH AND RECIPROCITY (MORALITY) So at NLI an

OUR PROGRESS WITH ALIGNING AI WITH TRUTH AND RECIPROCITY (MORALITY)
So at NLI and Runcible we are both creating a formal logic decomposer and compiler AND training an AI to make use of our logical system.
But strangely enough, If I gave you my system prompt, my user prompt, and the ten or so documents in my Chatgpt Project (“My Work”) you would find that it would produce the correct answer to nearly every question.
So without training, with just in-memory reasoning, ChatGPT can apply my work (our work) extensively
So it’s not that we have doubts that we have solved the problem of AI decidability. It’s that we want to ensure that it can produce the level of precision that we as humans do with edge cases.
So it’s not a question of success.
It’s just a matter of precision.
And that’s what we’re doing.

Cheers
CD