Part II: Intelligence and Implications

Introduction to Machine Intelligence

Outline for Today’s Lecture

Machine Intelligence: A Philosophical View

Information Search as a Pathway to Intelligence

Machine Intelligence

Humans are known to be intelligent. What makes us intelligent?

◼
  • ​
    • ◼
  • ​
    • ◼
  • ​
    • ◼
  • ​
    • ◼
  • ​
    • ◼
  • ​
    • ◼
  • Answers from previous semesters
    
    • ◼
  • Answers from others beyond the ECE101 class
    

    • Can machines also be intelligent?

    ◼
  • Yes
    • ◼
  • No
    • ◼
  • Undecided

    • As intelligent as humans?

    ◼
  • Yes
    • ◼
  • No

    • More intelligent than humans? Why or why not?

    ◼
  • Yes
    • ◼
  • No
    • ◼
  • Why do you think so?
    • ◼
  • ​
    • ◼
  • Answers from others:
    

    • Are the following “intelligent”? (Straight-forward cases)

    Calculator
    

    Your laptop
    

    Are the following “intelligent”? (Less straight-forward cases)

    Search engine--e.g. Google
    

    Roomba
    

    Rocket
    

    IBM Deep Blue that beat World Chess Champion
    

    Self driving car
    

    ChatGPT
    

    When Can We Say a Machine is Surely Intelligent?

    What would be an acceptable success metric?
    If you did not know who was behind the screen ...​

    The Turing Test

    ◼
  • Created by Alan Turing
    • ◼
  • The Imitation Game: A computer can be said to possess artificial intelligence if it can mimic human responses under specific conditions.
    • ◼
  • ​https://youtu.be/4VROUIAF2Do
    • ◼
  • The Turing Test was broken by ChatGPT: https://humsci.stanford.edu/feature/study-finds-chatgpts-latest-bot-behaves-humans-only-better
    • ◼
  • What questions can we ask to find out if it is a machine or a human?
    

    • Examples of what the Machine could not do (Until Recently)

    ◼
  • Captcha: https://support.google.com/a/answer/1217728​
    • ◼
  • Computers could not identify oddly written text.
    • ◼
  • But now we have "Captcha solvers"
    • ◼
  • Identify things in an image
    • In[]:=
    ImageIdentify
    
    In[]:=
    ImageContents
    
    ◼
  • What do you hear in this sound clip?
    • AudioIdentify
    00:00
    00:02
    Data in File[
    ....0.0/Data/Bird.wav
    ]
    

    What would you ask to find out if it was a human or not?

    ◼
  • ​
    • ◼
  • ​
    • ◼
  • ​
    • ◼
  • ​

    • What is Intelligence Then?

    ◼
  • Subject of much investigation.
    • ◼
  • Claim: Search is fundamental to intelligence

    • Best Tic Tac Toe move

    ​What is the best move for X?

    What is in this image?

    ◼
  • Cat or dog?
    • ◼
  • Object recognition has been an important part of computer vision for decades.
    • ◼
  • 2009—ImageNet: Image database of 14 million images for more than 21000 concepts (Original paper)
    • ◼
  • Best recognizers now use search, appear intelligent
    • If machines can solve search problems quickly, they can appear intelligent.

    Challenge for you

    Can you come up with a problem that can’t be solved by search?

    ◼
  • ​
    • ◼
  • ​
    • ◼
  • ​

    • Web Search

    So much content scattered over the edge of the Internet.
    ◼
  • Given a topic, how can I find the most relevant and interesting pages?
    • ◼
  • What if you don’t know they exist?
    • ◼
  • Search for them! Look everywhere!
    • ◼
  • Web search was the “killer app” for networks of workstations, a predecessor of the modern datacenter.
    • ◼
  • A company named Inktomi had the best technology (and most of the market),
    • ◼
  • But Google bought the company that invented the idea of selling ads for search responses.

    • How does web search work?

    ◼
  • Alice searches with phrase P
    • ◼
  • Search engine finds
    • ◼
  • a large set of scattered documents D
    • ◼
  • that contain P or are relevant to P:
    • ◼
  • D = {d1, d2, …, d203}
    • ◼
  • Search engine rank orders D by “relevance”:
    • ◼
  • D = {d17, d185, d23, d29, …, d12}
    • ◼
  • "relevance" could depend on the searcher
    • ◼
  • History of Nike (history professor) -> Greek goddess
    • ◼
  • History of Nike (political scientist) -> Nike missile
    • ◼
  • History of Nike (social media influencer) -> Nike shoes
    • ◼
  • Show results of search quickly

    • Wednesday: Anatomy of a Search Engine