Baal Zebul said:
why did i write it?
You need help? Well, it is pretty obvious that i wrote it to fit in the discussion, saying similar things to everyone else. But my pride extends further then the value of everyone else here so i added "Or Not".
Baal Zebul said:
Why you will never understand how every organic creature works is because you think to highly of yourself.
I'm sorry, I have simply no desire to discuss your (or my) personality.
Baal Zebul said:
It is not security related rules, it is a parameter that it can edit by itself...Not poor design, poorly raised.
I do not think you really know what you are talking about. Basically, good luck with the very deep analyses of all the system's input + impact on its current status and knowledge.. I think it's very practical to hardcode many rules in particular AI solutions. Self-modification limits are not necessarily limiting system's intellectual abilities and can prevent the system from many types of potential self-damage. BTW it's OK if such system develops powerful tools for its own use but the key configuration changes should stay well controlled by an authorized system/subject (ultimately by us). Our AI is an extension of our intelligence. Our intelligence is a tool we use to reach our goals. Our AI is simply supposed to work for us. If it's designed so that it can get reconfigured through regular input channels so that its goals conflict with our goals (let's say human goals) then what's the purpose? It wouldn't really be our AI any more. Why would we want to allow that? What would it buy to us? How would then be the development cost justified?
Baal Zebul said:
you will instead then have to cover that it is wrong to harm people directly or indirectly
And what would be the rule about "wrong" things? The fairy-tale-rule "Never do it!" or that "do it only when necessary in order to prevent things which are more wrong"? The key reasons why the criminal law is fuzzy include the fuzziness of the right/wrong and the more-wrong/less-wrong issue. Do you have a good sense of how many things do harm to us? You might get surprised if you research details about ingredients of products you use in your bathroom. Majority of this kind of products contain junk which is slowly killing us. Let's say you have a girlfriend. Even if
this story isn't exactly about her, she can hardly escape to similar type of harm. Research details about how our food is being processed and what are the reasons for all those procedures. Research details about vaccination, about product's used by your dentist, about chemicals in your mattress (and the reason why it's used), .. The list goes on and on.. There are known issues with so many widely used procedures and products that it's not easy for individuals to live a healthy life. In many cases, there are good alternatives for the product/decision makers but companies/corporations have their reasons why to ignore those opportunities, average people have no clue what's going on and many politicians make their dirty deals with those big corporations (+keep laws/rules which let them to do so kind of legally). You may even find that there is a law in your state which makes it more difficult for you to buy a particular lower-risk (or no-know-risk) products. This whole "harm" problem is kind of complex. It goes through many levels I did not mention. If ALL the harm is stopped in the near future then the world's economy would just collapse. It's not an option for us. I'm afraid we have a long harmful way to go and our AI systems will be required to choose from harmful options.
Baal Zebul said:
So, tell me how you believe that the human brain works (basic and general description), i managed to get it just onto a few entries in a process flow.
Key subsystems: senses, goals, memory, analogy detection, reasoning & planning, body control.
Key high-level goal: Happiness. Its level and stability defines the quality. Chemicals responsible for our happy feelings (the coolest stuff we got): serotonin, acetylcholine, noradrenaline, glutamate, enkephalins and endorphins. Let me know if I missed any. BTW the dopamine was relatively recently removed from that lovely family. the less-than-absolutely-happy feeling (provided by the system of senses) triggers intellectual procedure(s) in a healthy brain. If everything is functional, the information from senses is being continuously stored in memory. That information typically includes {A} the status of the body (the internal status info - that mix of all the current feelings) + {B} the current info (from senses) about the external world. that intellectual process which was triggered uses the analogy detection system to find the most similar "record" (/scenario) in the memory. The {A} is used like an index in a relational DB. It significantly limits the search-scope before the {B}-search is fully applied. That makes the search very fast but not perfect (because some of the {B} suff may not be there + another thing which I may point later if necessary). In certain scenarios, we are literally forced to miss some "records". When we have the top n closest scenarios, we check the "timeline" around those spots (in memory), using {B}-focused analogy search and trying to find a particular {A} which is at least a bit closer to the current-(sub)goal-related-{A}. If it's there then various types of reasoning are used to retrieve steps which could be possibly useful in the current scenario. Planning takes place and finally the body control tries to perform the steps. Results are continuously compared to the expected results and steps are regenerated if necessary. The brain is pretty good in finding something what's at least somehow similar (even if we do not really recognize the similarity on our current consciousness level) so there is always at least a tiny preference for the direction of thoughts/analyses. There are many important details I did not say. I do not think it's possible to describe this kind of stuff briefly and cover all the important stuff + I have better things to do now. ;-)