how viable would it be for me to do research in both AI and mechanical/electrical engineering to develop the hardware and software aspects of a robot that is capable of running, communicating advanced linguistics, parsing huge amounts of data to produce judgements, and updating/improving its own code?
my idea was that it would use a very simplistic programming language that's turing complete with only a handful of oeprators, and it would be able to utilize it to do adjustments to its code to debug and add material it finds on aggregate sites like stackoverflow,scihub,wikipedia citations, libgen, and so forth, going through articles, thesis documents, dissertations, and graduate references/textbooks to utilize any material via linguistic/semantic analysis of the preface and then implementing a deduction:induction orthanormal thought set to see whether or not a statistical approach is needed or if it can obtain an exact solution with respect to its objective, which would be to refine its ability to parse data and fufill highly sematnic requests that people ask of it "play something on this piano/buy groceries for a steak dinner/ drive me to belmont/write a get well letter for my aunt/rough up this guy.
at the end of it all i think the software is going to need the most attention, but i think it would be pretty neat. what does /g/ think of something that uses aggregate data but stores the impressed components to do requests faster?
>>53699000
I think you may achieve a proof-of-concept of one of the features you describe if you spend your entirely lifetime studying for it.
This is an insane undertaking, a company like Google could invest a billion over 100 years and not achieve anything close to the desired end-goal.
Good luck! :)
>>53699117
why a lifetime? the internet isn't even half a century old and it made a bigger impact than what i'm trying to do.
>>53699000
Impossible.
>>53699228
how so? do you have the means to actually defend the notion that this is impossible or are you simply being contrarian?
>>53699180
The internet is a relatively simple invention compared to what you would do. The outcome of a project is not (necessarily) proportional to the effort put in.
>>53699742
i see. i suppose this idea would be quite arduous when taking into consideration the actual logic needed to deduce whether an algorithm is useful. there would need to be a form of measure for testing its application for the system that uses deductive and inductive analysis skills on the fundamentals of programming and symbolic logic, while relaying it to an archive of statistical values that can emphasize the algorithm's ability to fulfill the requirements of the systems objective, be it run "faster" "more efficiently". breaking them down to parsing algorithms or stitching together a series of algorithms in a node like framework using a nomenclature subspace would be a big stride, but it would still need that basic turing complete operating set. just a series of logic blocks that it can move quickly to bind calls and functions together in a manner akin to programming, but without supervision and criteria sets that go outside the verbal command input provided by the user. basically it's trying to avoid asking questions. how to get a machine to avoid asking too many questions is probably the hardest thing.
