Exploring Intelligence and AI Development

Sep 4, 2024

Lecture Notes on Intelligence and AI

Introduction to Intelligence and AI

  • Intelligence has been a key problem for humanity over 2000 years.
  • The term "artificial intelligence" was coined at a conference in 1962 by John McCarthy and Marvin Minsky.
  • AI has seen significant progress, especially in the last 20 years, driven primarily by computer science and common sense.

Importance of Interdisciplinary Collaboration

  • Understanding intelligence requires expertise beyond just computer science.
  • Collaboration involves disciplines like:
    • Neuroscience
    • Cognitive Science
    • Various institutions (e.g., MIT and Harvard)

Milestones in AI Development

  1. Deep Blue: IBM's chess-playing computer that defeated Garry Kasparov.
  2. Watson: Beat champions in the quiz show Jeopardy.
  3. Drones: Demonstrated ability to land on aircraft carriers.
  4. DeepMind's AI: Learned to play 49 classical Atari games better than humans.
  5. Mobile AI Systems: Capable of visual recognition and responding to surroundings.

Current State of AI

  • The field is experiencing a golden age for intelligent applications.
  • Great opportunities for innovation and profit in AI.
  • However, there's a gap in understanding how humans interact with images and process visual information.

Research Focus Areas

  • Investigating how the brain answers questions about images:
    • What is present in the image?
    • Actions or thoughts of individuals depicted?
    • Storytelling based on visual input.
  • Aim: Develop systems that mimic human cognitive processes while answering these questions.

Turing Test and Beyond

  • The Turing test is not sufficient; researchers aim for a deeper understanding of intelligence.
  • Turing Plus Plus Questions: An exploration of intelligence that goes beyond simply passing the Turing test.

Defining Human Intelligence

  • There is no universally accepted definition of intelligence due to its diverse forms.
  • Focus: Understanding human intelligence through:
    • Neuroscience
    • Computational models
    • Psychophysical responses

Research Directions

  • Current focus at the center is on vision-related tasks.
  • Goals include:
    • Understanding computational, psychophysical, and neural levels of visual processing.
    • Developing models for face recognition:
      • Utilizing findings from monkey studies and human fMRI to inform AI models.
  • Challenges remain in studying more complex cognitive tasks like storytelling.

Conclusion

  • Research in AI and intelligence is ambitious and multifaceted.
  • Continued interdisciplinary collaboration is essential to advance understanding and technology.