Timothy Shreffler

Linux Systems Administrator, Booz Allen Hamilton

This conversation is closed.

We should create adaptive systems to support humanity in it's goals.

What if, we created a server/os/program that could create virtual machines on it's own. The administrator first creates a virtual machine template with a purpose (mail server, apache server, etc.) We then have a team of developers take a systems administrator wiki and start automating everything by writing a program that can do it. So the server would be adaptive, it would act upon input much like a human. The server creates the virtual server, the virtual server uses programs to automate tasks. If we focused on this we could probably create an endless loop, slightly scary because I think that was in the AI movie. Just a thought...

  • thumb
    Jan 27 2012: The problem is that we can only write program's for known situations. As programming uses an "if ... then ..." mechanism it's hard, if not impossible as far as I know, to write a program that adapts to unprogrammed situations. That's why we will probably always need humans to operate the systems and can't have a fully self-sufficient program.
    • thumb
      Jan 27 2012: Thanks for chiming in man... Two thoughts come from your answer. The first being maybe it's a better idea for us not to have a self-sufficient program as I saw what happened in iRobot LOL. The second would be more of a question as you certainly know more about this then I do. If we wrote enough if and then would we eventually make a loop? Much like if you click the first link in a Wikipedia article over and over again eventually you will end up at Philosophy no matter what the subject of origin is. If we have a basis a foundation the possibility of the computer not solving the problem the first time around is possible but if it loops we get a second chance. The next question being would the computer come to the same conclusion over and over again? Maybe we should test it, see how it works, and then see if we could find out where the disconnect is? What do you think?
      • thumb
        Jan 27 2012: You should read the book Robocalypse if you're interested in that kind of fiction!

        I'm a paid freelance programmer next to my Information Science. I'm not infallible though.

        I think it's possible to create systems that learn from situations automatically by their results and can predict the outcome of certain situations based on results from previous (similar) situations. Of course you could channel any problems the system hasn't occured before on to experts so they can solve them and the system can learn from it, but then it wouldn't be a self-sufficient-system but only decrease the need for experts. You can read more about these kind of systems here or Google for the term "Expert Systems":

        To answer your second question: Yes, the result would always be the same. Saying this I am assuming that the input the system receives (from the outside, like the situation, and from the inside, as in the database) would be precisely the same as the system uses math to calculate the solution: A calculation always has the same result, untill you change a parameter.
        • thumb
          Jan 27 2012: Ok trying to wrap my head around this while working at my job hahaha... If we created this loop which essentially would always have the same outcome. What if we put in a always changing variable? so the outcome was never the same... granted this won't be the best outcome... But if we then set up a purpose for it we can take the outcome which is never the same and use some if and then statements to direct it. Essentially like learning we would have to work through each (always changing variable) at a time. Record that choice in a database and have it pull from there like past experiences. I suppose the harder part would be teaching it how to relate things to one and other... I feel like this is possible hahaha I just don't know how. The good thing is you never know what will happen, our conversation here on TED could lead to something amazing.
      • thumb
        Jan 27 2012: But realise that your always changing variable that you want to try each possibility for and record it is not simply a number, but an attribute of a situation. Let me draw a situation for you:
        We program a system that calculates whether firefighters are allowed to enter a building. We give the system as input possibilities:
        - The number of firefighters / victims
        - The size of the fire
        - The complexity of the building
        Then, after implementing the system, a fire occurs:
        - There's 3 firefighters and 2 people inside, unconscious
        - It's a small fire
        - The room with the people is easily accessible
        But there is also a gas leak about to happen, as the pipe is of bad quality and likely to burst. (Theoretically, probably not likely in real life, but I dont know anything about gas pipe's..)
        The system wouldn't be able to make a right decision because we can't give it all the information we need to. An expert would have to make a quick decision due to the time limit of this situation and the system needs updated.

        Of course this is not very likely a thing to forget to build into the system but it's an example: We can never predict every possible attribute of a situation. Hence we will always need experts for when the system doesnt suffice and the system will never be self-sufficient. The risk of assuming creating a self-sufficient system is possible is that we will end up having no experts when the system can't calculate the right answer. That's why we should always remember we need to be able to replace systems like these with humans may the need occur.
        • thumb
          Jan 27 2012: That's a good enough answer for me, I can accept the fact that human beings will forever be smarter then computers. LOL actually I'd prefer it to be that way as I am a human :-)
  • thumb
    Jan 27 2012: If we could write choice as a variable I wonder how that would be done... It seems if and then are not enough :-) or maybe right why as a variable? Curious are all variables constant? Could we have a variable that never stays the same in the equation therefore always creating a new answer? Maybe call it the change variable... I'm sure everyone that writes programs has thought of this already LOL
    • thumb
      Jan 27 2012: You could write a variable that's random with one simple function, but I assume you want the solution to be the best one and not a random one? This would have to be done with a mathematical function by calculating the chance of success per option. The problem is that the system has to give a certain value to certain "features" of a situation which have to be predetermined or learned based on history.
      See a system like this as an automation of the human mind when it makes a choice: Before we can write a system that can make decisions like humans do we have to know precisely how a human brain calculates the best option and implement this in a mathematical way. If we can accomplish this the next step would be to give it more experience than a human brain by feeding it as much input, based on historic facts, as we can.