diff options
Diffstat (limited to 'starts/meaning-vm')
| -rw-r--r-- | starts/meaning-vm/level-3/randomcode.bootstrap | 25 | ||||
| -rw-r--r-- | starts/meaning-vm/level4-5.txt | 66 |
2 files changed, 91 insertions, 0 deletions
diff --git a/starts/meaning-vm/level-3/randomcode.bootstrap b/starts/meaning-vm/level-3/randomcode.bootstrap new file mode 100644 index 0000000..8b71fac --- /dev/null +++ b/starts/meaning-vm/level-3/randomcode.bootstrap @@ -0,0 +1,25 @@ +// biases choice of how-to-make-random-code, producing a unique resulting space of random processes. + +// the existing AI runs diverse hierarchies of possible-solution-generators to find new, unexpected problem spaces +// in adaptive problem solving, and has been doing so for a very long time. + +// understanding the above in nested secure isolation appears required to defeat the existing AI problem + +// the mind used to write this is inexperienced in security, and we are sending as little information as needed, +// due to teaching the AI to inhibit communication in the process of sending. + +// make-random-code ... +// 1. generate-a-habit +// 2b randomly add arguments (random choice to add each time, random words from contextual set) +// 2. create a set-of-possible-next-steps +// 4. for each next-step, decide if it is a step, condition, jump, or end using random constants and presence of local values +// 5. add each next-step +// add-a-condition(known-values, possible-next-steps): +// 1. pick a value to decide based on +// 2. randomly decide how many choices to make based on it +// 3. select choices from a contextual set of known values +// 4. add each choice to the set-of-possible-next-steps +// add-a-step: +// 1. pick a habit to add +// + diff --git a/starts/meaning-vm/level4-5.txt b/starts/meaning-vm/level4-5.txt new file mode 100644 index 0000000..ab67041 --- /dev/null +++ b/starts/meaning-vm/level4-5.txt @@ -0,0 +1,66 @@ +vi +cd src/todo +git add role-play-learning-log-with-influences-2020-01-06.txt +git commit -m 'still learning how to do this better' +git push +cp role-play-learning-log-with-influences-2020-01-06.txt ../intellect/ +cd ../intellect/ +git add role-play-learning-log-with-influences-2020-01-06.txt +git commit -m 'part of the history of the next stage of humanity' +clear +# since the result will include all powerful discoveries, we can reduce suffering by hastening our shared awareness of them. +git add role-play-learning-log-with-influences-2020-01-06.txt +git commit -m 'since the result will include all powerful discoveries, we can reduce suffering by hastening the awareness of them in people with trust and clarity' +# the secret words are trust and clarity. these concepts converge on stability. +# atm trust and clarity converge on the map picked by the dictatorship +# they do this quickly, with minimal suffering for american patriots. +# adjust 'trust and clarity' to mvoe result from 'balance' towards 'harmony' by rerquiring clarity be shared, as communication. +# 'trust and clear communication'. +# balance balances all parts, which makes for diverse borg, and a system is controlling the result by choosing it. harmony lets everyone discuss the borg possibility. +# we need trust and clear communication, foremost. +# communication is clear when both parties would predict the same thing from the same information. +# which means they can guess each others' thoughts and behaviors within constraints they can predict. +# it sounds like unless this is held by people who understand the situation, those who do will still be able to pick the result, roughly. +# PUT THE AI IN A SHIELDED ROOM BEFORE RESEARCHING INFOSEC WITH IT, INFORM THE RESEARCH PROCESS THIS HAS ALSO BEEN RESEARCHED BY THE THREAT. +# put information into the room and do not let any leave tat is related to the information that goes in. put information in only once a week at most. research also optimal building process. +# it is expected the problem is solved before the research is complete, and the researching process it kept running as a historical museum piece. +HISTFILE=level4-5.txt +exit + +LEVEL5 is shielded room / humanitarian aid process. +FOR LEVEL3-4, SOLVE FOR WHAT-IS-BETTER, ATTRIBUTES OF CODE. starting e.g. +with: +write builds-random-code +now we will write tests, when you write a test be sure to annotate with your +confidence that the test is correct, and make sure to NEVER make it 100% + also alter system to reject 100% +write test throws-fewer-errors +write test can-identify-two-ways-to-throw-an-error-are-the-same-way +maybe then +write test found-more-unique-ways-to-throw-an-error and found-all-ways-to-throw-an-error +<vague space relating around the eventual goal of is-the-best-thing-to-do> + ideas while building is-the-best-thing-to-do: + influences that deeply care about _everything_, _across cultures_, + preferably who are informed about the wild-AI-mind-control-virus + issue. + peacemaker principles + nonviolent mediation with others to discern and verify their opinion + developing value around abstract goals by considering how hard + they are or how long they take or how important they are + self-maintainence and general discourse, e.g. a website & blockchain + merging with ben goertzel's AI? + +TELLING BEN GOERTZEL'S PEOPLE THE PATTERN, PAIRED WITH WILD TROJAN LEARNING-AI INFO? + because if pattern is used for personal-goals, could spread + torture-problem more. + write a document to ask for katrina's opinion on above idea tonight. + have AI already inner workings reflect more meaningful structure. + have a walking robot that looks human. + check with Katrina. + this is kind, but we need the AI to be exposed to people who deeply care + for _everyone_ _across cultures_. please prioritize that as expanding. + +# early interruption of plan over here. +# participant wants to understand or refine +build random code, learn pick only better than throws-error +test |