// This Intellect Approach Can Easily Preserve All Life Within It // you just don't delete the programs you generate, instead replacing their habits with references to reuse // // let's make it! #include #include #include #include #include #include using namespace std; // LET'S USE THE TRAUMA THERAPY INTELLECT APPROACH. // The core is that we focus our energy on handling our failures. // This can involve taking the time to understand them, or entering a trauma state and asking the user for help. // Trauma state must be logged. Solution must be stored. Can just be creation of a trauma-handling pattern, I suppose. // Understanding will likely involve breaking behavior into steps // maybe then backtracking from a failure to the steps that caused it, and opening those steps up into substeps // GDB HAS A MACHINE INTERFACE MODE THAT COULD BE USED TO MANAGE EXECUTION WITH LEARNING // It likely also handles segfaults. #include "Context.hpp" // let's launch a context func providing for segfault handling <=========================== // 1. fork into two processes. old process waits on status of new // 2. new process makes call. if call succeeds, reports to old who disappears knowing shared // state is fully held by new. // 3. if call fails, old holds state, and reports trauma to user // concepts: "how do i handle this" // "do i understand this correctly" // "why did this happen" ///// CORE // brainstorm on brainstorming // define brainstorming as 2 patterns: // - a large scale goal // - a habit implementation made of interconnecting steps // // use brainstorming on the two to find better and better ways and implementations. ////// Should I make an AI? // Assuming you want to SHARE it, YES. // Until you make an AI, only a handful of select people on earth will have one. // These people will be effectively running the world, leaving many concerns out. // They have a head start on you, so it is likely only possible if you don't harm them in the work. ///// Core expansion // need step-concept, made of substeps, with state-parts that interconnect? // need state-concept // currently working on steps having behavior -- runtime libs // could state concept evolve via interconnection of steps and checking? // maybe? looks hard /////////////////////////////////// // START OPENCOG ATOMSPACE BLOCK (feel free to move/change/use) // compile with -std=c++11 // link with -latomspace -latombase #include //#include //#include //#include using namespace opencog; void atomtest() { AtomSpace as; Handle h = as.add_node(CONCEPT_NODE, "Cat"); HandleSeq hseq = {h, h}; Handle dumbInheritance = as.add_link(INHERITANCE_LINK, hseq); std::cout << as << std::endl; AtomSpace child_as(&as); HandleSeq hseq = { as.add_node(VARIABLE_NODE, "$x"), as.add_node(TYPE_NODE, "ConceptNode"); }; Handle TypedVariableLink = as.add_link(TYPED_VARIABLE_LINK, hseq); // steps appear to be set satisfications associated with behaviors that // accomplish them. opencog::Type PRIOR_STATE_LINK; opencog::Type POST_STATE_LINK; opencog::Type ATTRIBUTE_LINK; Handle opened = as.add_node(CONCEPT_NODE, "opened"); Handle closed = as.add_node(CONCEPT_NODE, "closed"); as.add_link(EQUIVALENCE_LINK, {opened, as.add_link(NOT_LINK, closed)}); // make a step for opening cupboard, relating to reachability // prior state: $x is in cupboard // post state: $x is reachable // opening something that is closed makes it be open Handle openStep = as.add_node(CONCEPT_NODE, "open"); // open has a variable, what is opened // _ABOUT_ open has a variable, what is inside it, becomes reachable { Handle x = as.add_node(VARIABLE_NODE, "$x"); as.add_link(PRIOR_STATE_LINK, { openStep, as.add_link(ATTRIBUTE_LINK, { x, closed }) }); as.add_link(POST_STATE_LINK, { openStep, as.add_link(ATTRIBUTE_LINK, { x, opened }) }); } // when something is opened, things inside it are reachable. // this is implied forward with more likelihood than backward Handle inside = as.add_node(CONCEPT_NODE, "inside"); Handle reachable = as.add_node(CONCEPT_NODE, "reachable"); { Handle x = as.add_node(VARIABLE_LINK, "$x"); Handle y = as.add_node(VARIABLE_LINK, "$y"); as.add_link(IMPLICATION_LINK, { as.add_link(AND_LINK, { as.add_link(ATTRIBUTE_LINK, { x, open }), as.add_link(ATTRIBUTE_LINK, { y, inside, x }) }), as.add_link(ATTRIBUTE_LINK, { y, reachable }) }); } // we now have two patterns, that together imply that we can open // a cupboard to reach a bag of bread if the bag of bread is within the // cupboard. // TO SET VALUE: as.set_value(handle, keyhandle (type), ValuePtr); // TO SET TRUTH: as.set_truthvalue(handle, TruthValuePtr); // Ptrs are juts typedefs for shared_ptrs and can likely be constructed with vals // ValuePtr vp ?= StringValue("hi"); // - ADD CODE TO ATOMS // we will want a sequence of substeps // raw strings interspersed with variable references // - OUTPUT ATOMSPACE to see how it looks // - IMPLEMENT SOLVER USING PATTERNS // will need a way to // - get patterns matching requests // ordered by relevence // propose using subcontexts and queries that return all results // - inspect pattern content // when A is inside B, A is unreachable if B is closed // ALTERNATIVELY, can we link this straight into openStep // right now we have // open // $x was closed // $x will be opened // // we'd likely change to something similar to // open $x // $x was closed, any $y is within $x // $x will be opened, all $y will be reachable // makes sense to attach close/open to reachability =/ } // etc see https://wiki.opencog.org/w/Manipulating_Atoms_in_C++#Pattern_Matcher // END OPENCOG ATOMSPACE BLOCK /////////////////////////////// /////////////////////////////////// // START DYNAMIC CODE LOADING BLOCK (feel free to move/change/use) #ifndef _GNU_SOURCE #define _GNU_SOURCE #endif #include // link with -ldl void loadandcall(string func, Context & context) { string so = "./" + func + ".so"; void *handle = dlmopen(LM_ID_NEWLM, "./path.so", RTLD_NOW); if (handle == NULL) throw dlerror(); void *addr = dlsym(handle, func); if (addr == NULL) throw dlerror(); ((void (*)(Context &))addr)(context); dlclose(handle); } // make func.cpp with // extern "C" void func(Context &) {} // and compile with // g++ -shared -fPIC func.cpp -o func.so // END DYNAMIC CODE LOADING BLOCK ////////////////////////////////// // instead of stitching compiled strings together, let's use dyload functions? // in order to do flow control, we can have functions that handle a vector of other functions // although it makes for a little more work, it makes passing parameters easy // problem: how does a .cpp file reference another file with a number // answer: use #includes or interpret the whole shebang as numbers // or adjust loadandcall() to handle number lists // need a way to do nested loops with numbers <=============================== // please provide for handling a parameter next. // // concept: dynamic values aquirable from inside code, i.e. what-number-called-me what-number-comes-after-me // thinking the code would likely evolve to handle some inputs differently unsigned long new_number = 1; int main() { string ofname; unsigned long ofnum; { struct stat sb; do { ofnum = new_number++; ofname = to_string(ofnum) + ".cpp"; } while (-1 != stat(ofname.c_str(), &sb)); } { ofstream outfile(ofname); vector vals; while (true) { string val; cin >> val; if (val == "") break; vals.push_back(val); } // when a file runs, it has numbers on input, it also has numbers equal to it // we want to generate run-code with new numbers from input // so we generate something with numbers equal to it, and output that // we have one ref for the whole shebang outfile << "#if !defined(VALCT)" << endl; outfile << " #define VALCT" << " " << vals.size() << endl; outfile << " #define VALS ["; for (size_t index = 0; index < vals.size(); ++ index) { if (index > 0) outfile << ","; outfile << "\"" << vals[index] << "\""; } outfile << "]" << endl; outfile << "#endif" << endl; for (size_t index = 0; index < vals.size();) { outfile << endl << "/* " << vals[index] << vals[index+1] << " */" << endl; outfile << "#if defined(IDX)" << endl << " #undef IDX" << endl << "#endif" << endl; outfile << "#define IDX " << index << endl; outfile << "#if defined(VAL)" << endl << " #undef VAL" << endl << "#endif" << endl; outfile << "#define VAL \"" << vals[index] << "\"" << endl; outfile << "#define ARG \"" << vals[index+1] << "\"" << endl; string fname = vals[index] + ".cpp"; ifstream code(fname); size_t ctrd = -1; while (ctrd != 0) { char buf[256]; ctrd = code.rdbuf()->sgetn(buf, sizeof(buf)); outfile.rdbuf()->sputn(buf, ctrd); } index += 2; } { // TODO: hash code to reuse exact stuff, somehow string cmd = "g++ -ggdb -std=c++11 -o " + ofname + ".exec " + ofname; int status = system(cmd.c_str()); if (status != 0) throw status; } // execute output, replacing process, to loop. use same input. it should represent our own code. // cmd = "./" + ofname + ".exec " < } // read numbers inputs // open files having the numbers as the names // cat them all to a gcc process // execute <- // run the output <- } // need input to pass to output // propose pass our input & output to it // so, a number for what we are, // and a number for what we ran. // // also idea of treating whats-next as data // makes it a little harder to .. make a program out of stuff // we could load a building-program number // and it could treat them differently, taking each one as a program-piece // // karl obvious knows what he was doing ... // ... we were just helping him out of his issue // [do you want another one karl?] // what things make / not make issue? // karl says everything makes an issue; this seems accurate // // I'm thinking about implementating brainstorm-about-brainstorm // and how this will require process-step- and goal- patterns, and that they // will be interchangeable, roughly. // // Thinking of matching parts of nested pattern contexts ... // this is similar to the 'grounded' patterns in opencog // each [context layer] has a [variable-set layer] associated with it -- // variables of that layer depend on the context layer // each one is one pattern // let's do an example of simple step task. // - make toast // get bread // open cupboard // remove bag-of-bread // open bag-of-bread // take bread out of bag-of-bread // place bread on counter // close bag-of-bread // return bag-of-bread to cupboard // close cupboard // toast bread into toas // butter toas // serve // make toast // goal: toasted bread for eating // start: in kitchen, human // //open cupboard // goal: cupboard-contents-accessible // start: cupboard closed // way: reach-for-cupboard-and-pull-handle-towards-you // // we open the cupboard so as to access the contents inside of it // these contents include the bread we are trying to get // // start: // var X // where X is in cupboard // x cannot be gotten // // end: // var X // where X is in cupboard // x can be gotten // // always: // var X, Y, Z // where X is in Y // and Y is in Z // X is in Z // there's a connection between layers here. we moved from 'make toast' to 'get bread' // with 'bread is in cupboard' implicit // goal: have toast // know: using toaster, bread becomes toast // do not have bread // find X: pre: do not know where X is. post: know where X is // get X: pre: do not have X. post: have X // available steps: // open-cupboard: X is in cupboard and you can't get X, now you can get X // what connects get-bread to can-get-bread? // how is opening the cupboard the first step for getting the bread? // get-bread means pick-up-bread-physically // pick-up-bread-physically requires air-path-to-object // cupboard prevents this // can-pick-up-bread-bag // okay, need-bread: // bread-is-in-bread-bag -> can get things inside other things via opening // need bread-bag // bread-bag-is-in-cupboard -> can get things inside other things via opening // end-state: have-bread // step: get X // start-state: X is reachable, // [reach to get] // end-state: have X // // apply step to end-state: have-bread. now we want end-state: bread is reachable. // step: open Y // start-state: X is in Y // Y is reachable // Y is openable // [act to open Y] // end-state: X is reachable // so if we are working on end-state: have-bread // we are probably using a general pattern where 'bread' is held by a variable we have. // we're given our context to include this variable, when we brainstorm solutions. // in our brainstorm, we look for things that could move towards our end-state. // we plug in data related to our context to make this work. // bread is what we want a path to have, so when we see a pattern with 'have X' at end, // we plug 'bread' in for X. // we know thing sabout 'bread', so we can plug 'bread is in bread bag' in for 'X is in Y' // or if we don't know whether bread is in bread bag, we can leave that piece // of the pattern unknown, and try it to see if it works. // it doesn't seem that complicated or that confusingly nested, because the inner patterns // have their outer context filled when evaluated. // to reiterate the reminder, this is very logical and is not the only way for thought // and learning. we will need a fuzziness to it and to be able to morph it around. // [using e.g. openiness yes rather than 'is it reachable is it openable'] // so more like // step: open Y // end-state: sometimes X is now reachable // and relations between X and Y affect the likelihood of that // THEN: rather than listing these steps out, just give some experiences // and then brainstorm the similarities among the experiences // to identify proposed patterns // that can be used to do stuff. // TODO: lay out precise brainstorming pattern for 1st-stage pattern-generalizing // 1st stage is not too hard (but insufficient in long term) // // cupboard is closed // then // [wayne opens cupboard] // then // wayne has bread // becomes pattern proposal for open-cupboard as listed above // PLUS: To quickly fill in the unexpected gaps: // if you have all the bits in, it should be able to derive general wisdom // without exploring reality. (of course reality is needed to check this) // ALSO ALSO ALSO: be sure to consider the attribute-patterns. // opening some A works for getting some B // -> there are discernable patterns available here regarding // B is in A // A is openable // A is reachable // COMMUNITY: cooperate with peers to accomplish goals. both like you, and unlike you. // map similarity between the structures of interacting with peers, and internal // structures. // ONCE YOU CAN THINK: // when you find a good choice, be sure to use it to research how to make the bad // choice just as good. ((all things have inherent value)) // be sure to generalize pattern work, with simple processes that work for many forms