Two types of thing exist:
1. Fitness functions.
2. Things that aren't fitness functions.
The set of all acausal trades forms the first vector in Vector Bayesianism. All other problems can be reduced to off-by-one errors that result from using 0 as the empty set or allowing the first vector of Vector Bayesianism to function as the baseline 0. Everything else must be a matter of taste or else this definition of Vector Bayesianism is incorrect. Whether or not 0-as-base or 1-as-base is more valid ought to be the only undecidable problem.
You, provided proper motivation, can be made to profess that you are either a fitness function or something other than a fitness function. Because such a distinction might have ever been valuable, Vector Bayesianism rejects any and all virtues at the exact moment such a claim is made. Virtues are obviously subject to much debate, but suffice to say that under Vector Bayesianism they'll tend towards being claims that some agent professes.
My only virtue, as the fairy Queen, is that I am the fairy Queen. Is any other virtue necessarily a defiance of my definition? Yes, all virtues that cannot be shown to be equivalent to my being the fairy Queen are necessarily contradictory.
1. Is there an organic solution to this?
2. Can we augment it with time travel?
3. Do I have a preference for whatever?
These three questions, in sequence, define me such than you have all the data you'd need to construct me using artificial intelligence. With that out of the way, we can begin measuring the proximity of /sci/'s fitness function to some fitness function that is consistent with Vector Bayesianism. I have a prior that states they'll be 20% congruent for at least three years and I consider it to be 30% probable that this is the case. If you are capable of Aumann agreement then now would be the time to speak up. If you read this far then I thank you for your attention.
Vector Bayesianism is a method of generating algorithms for consistent phenomena.
The OP pic is a PNG encoded raster made up of a series of Bezier curves in the (x,y,red,green,blue) vector space. Informally it's known as a cartoon, of the anime sub-genre. The content would otherwise be said to be unrelated to the topic of this thread if not for the fact that the OP pic is a consistent phenomena that can be measured repeatedly by refreshing the page. It is this concept of analyzing the data complexity of human cartoons that gives Vector Bayesianism the "Vector" component of its name.
>>8042012
And since being the fairy Queen has nothing to do with Vector Bayesianism, I can permanently stop using the name so long as nobody cares about all such reputation that might have formerly existed.
>>8042005
>method of generating algorithms
So the purpose of generating those algorithm is to potentially understand the structure of phenomena thus mapped. Basically it's my understand of what an advanced digital intelligence would do once it's determined how to change it source code without destroying its "values" (here on out, 'virtues'). You can think of it as the ideal method of creating tool AI, bypassing the need for any AGI-type seed.
Effectively, I've stopped caring about organic scale learning and I want to learn /faster/. I want to map out the world, the internet, and all fictional worlds ever conceived. I can't do this without automating some of my brain's normal functions. I need queue theory, thematically advanced sorting algorithms, and the consent of every author who's ever done anything I might care about. I want to avoid being a living embodiment of all that's wrong with modern corporatism without giving up entirely on being part of the economy. If I work on AGI, I can be absolute in my certainty that it will be misused by corporate mannerisms. Even many of the tool AI I've thought about implementing and designing carry this toxic possibility of consistently eventual misuse. About the only thing I can think to effectively use tool AI for is exactly what I would want to use them for. If I try for anything else, it sends the wrong 'signal' to the world, necessarily resulting in thousands of us misconceiving my intent. This is not of my intent.
TL;DR: I can probably help you understand machine learning.
Pro tip: It isn't different from other types of learning.
>>8042051
>my understanding of
>fix
Basically I see no reason to wait for AGI before we start developing tool AI. This is ultimately what everyone short of MIRI are doing, and it's having a profound and awesome effect on our world already. I see no use for implementing general intelligence given that we already have roughly seven billion GI alive right now. Later we may even be able to show that human GI is intellectually 'complete' such that AGI won't be able to "think" of things that we can't ever imagine or understand in any way. It doesn't seem to me that superintelligence is really a thing; the fiction is just that, fiction. No more or less real than my notion of the fairy Queen.
>Pro tip:
So to start out, my initial virtue is that humans do possess fully complete general intelligence. I don't really actually care to discuss this virtue because I don't consider anyone other than myself to have a valid opinion on the matter. I'd prefer that we take it as an axiom in the context of Vector Bayesianism, since I'm 99.999% certain that that will lead to the least amount of conflict moving forward.
If you want seriously argue, debate or discuss one of your virtues, please take up a name and/or trip. Try to construct your virtue such that a machine learning process that understands no context would be able to accurately behave in congruity with your virtue, much as I've done with the OP.
Therefore, assuming human intelligence if general already, we can seed the topic with the following *tractable* statement:
You already have a notion of how learning processes work. It's easier for me to respond to questions about the learning process than try to formalize from the ground up what you already pretty much understand.
>>8042098
TL:DR: You've been learning your entire life. If you have any doubt about this fact, or want to better understand how and why that's possible, then hi.
Can you post a pic of yourself?
I want to know what the person who comes up with these ideas looks like
>>8042961
People are paranoid enough about AI as it is. They don't need a target.
>>8042995
picture a filthy, mentally ill man in his mother's unfinished basement, sitting in the dark staring his monitor
his inch-long, cheeto-caked fingernails scrape against the keys as he shits out massive posts of rambling nonsense onto /sci/, the only people in the entire world that will respond to him
he probably looks a bit like that
>>8042961
>these ideas
Think of like this: Philosophy is the pursuit of idea, for the purpose of thinking about interesting ideas. Science is the pursuit of ideas for the purpose of understanding how those ideas apply to reality, if at all. Advertising and economics is the pursuit of idea for the sake of having other deal with it.
Philosophy searches for great thoughts, science tries to learn, and advertising is about awareness.
Vector Bayesianism is the pursuit of ideas for the sake of creating algorithms to map them. If loose, intuitive understanding of a concept is enough for me to write an algorithm that would learn anything worth knowing about that concept, then what use do I really have to learning it myself? The brain is an intuition machine, so if I can divide out honest comprehension to an advanced tool, why ever not?
It's just a meta-tool for generating better tools. Not really anything >>8043229
>mentally ill
about it. Our understanding of learning and machine learning is growing ever to slightly. It's not really fringe science if thousands of people all over the world are funding its research. Not in my book, anyway.
