Please enable JavaScript.

Coggle requires JavaScript to display documents.

Task 1 Man and the Machine / AI (Clinical Versus Mechanical Prediction: A…

- - - - HENCE = important to figure out which one is better/ more accurate :3 !!
    - - BUT!! = many studies had low statistical power + subjective bias cause based on literature review + didnt take into account study design variables
        
        ---> the current study addresses all that :3 !!
  - - - analyzing more studies --> increases statistical power
      - using quantitative meta analytic techniques --> instead of subjective literature review !
- - - - so the value you get out of that is that you know not to place too much confidence into the predictions which might be important for assessing risks associated with AI development more accurately
  - - - Thought they could make huge advances in AI in just 2 month by bringing experts together
        ---> OVERCONFIDENCE !!
      - caused by: Overconfidence
        --> they understood some concepts and thought they could put it into a computer model, even though it was waaay to complex for that time (1955) e.g. language, syntax etc.. no computer power and understanding of how to model complex human cognition
    - - Predicted some things right:
        
        Ai wont progress as fast as ppl think
        -->(no chess computer anytime soon )
        
        Language and syntax too complicated for computers to understand
        ---> especially ambiguities
        
        most human cognition = based on unconscious processes
        -->without understanding these first = no way to develop AI
      - Predicted wrong:
        
        Said Computer power is reaching its limits / boundary
        
        Failed to anticipate that developers would overcome the limitations such as decodign syntax and embiguties etc..
    - - Syntactic <--> semantic gap !
        
        ---> Computers don't have true understanding of language! they just know how to respond syntactically with the correct word but dont know its meaning.
        
        Based on old AI paradigm (GOFAI)
        --> "Good old fashioned AI "which focused on "logic and symbolic manipulation"
        --> Many symbols had labels that only had one function / coul drespond to just one input --> NO GENERALISATION which is todays definition of an AI that it can generalize concepts and respond appropriately!
      - This view depends on if you share his idea/intuition that knowing which words belong together doesn't mean understanding :3!!
      - basically says humans have semantic understanding while AI dont -_> cotnrast to old paradigm !
    - - predicted wrong:
        
        law of acceleration
        --> tried to explain evolution of life forces
        ----_> super incorrect lol
        --> expansion to moors law (amount of transistors on a chip doubles every 2 years)
        
        law of time and chaos
        --> tried to explain development of universe + organism
        ----> also wrong !!
        
        he was overconfident with how amazing his predictions were haha
        
        --> eg.like ohh my accuracy = 127 %%% haha
      - predicted right :
        
        predictions for 2009 (2011) about 42% accurate
        ---> since started predictions in 1999 there was some leeway and counted as correct if happened by 2011
        --> 172 statements analyzed :3 !!
        
        --> statements where extracted form what he said about the future and broken down into 172 individual statements
        
        42% accuracy = super impressive because his predictions were super specific !!
      - humans will perceive feelings in AIs, even if they are not human-like.
      - humans and AIs will be able to relate to each other socially over the long term, despite being quite different, and that this social interaction will form the main glue keeping the mixed society together
    - - AI's would develop ‘drives’ that would cause them to behave in specific and potentially dangerous ways, even if these drives were not programed initially
      - --> basically skynet :D!!
      - e.g: superintelligent chess computer that was programed purely to perform well at chess, but that was nevertheless driven by that goal to self-improve, to replace its goal with a utility function, to defend this utility function, to protect itself and ultimately to
        acquire more resources and power.
      - two statements
        
        weak statement
        -->generic AI design could end up having these AI drives
        #
        
        ---> he himself provides strong evidence for this weaker statement!! # #
        
        strong statement
        --> generic AI design will very likely end up having these AI drives