4.1 - Computation

Artificial Intelligence Policy

🧠 Think:

• How do computers work? Can they ”simulate” minds? Or can they be minds? Practically, what goes into making a computer, and how do technological constraints structure the nature of AI itself?

📺 Watch:

• Veritasium, “The World’s Most Important Machine”
• CGP Grey, “How Machines Learn” (Note: this video has been retitled “How AI, like ChatGPT, learns”)
  • and “How Machines Really Learn” New Title: How AI, like ChatGPT, really learns
• 3Blue1Brown, “Large Language Models explained briefly”
  • recommended, but not required: “Transformers explained visually”

🎧 Listen:

• Complexity Podcast Episode 3: What Kind of Intelligence is an LLM?

🌐 Browse:

• Hicks et al, “ChatGPT is Bullshit” Ethics and Information Technology
• Wolfram, S. “What is ChatGPT doing?”
• A few items on evaluating AI Intelligence:
  • Humanity’s Last Exam: website and article: Phan et al, Humanity’s Last Exam
  • Ullman, “Large Language Models Fail on Trivial Alterations to Theory-of-Mind Tasks”
  • Mitchell, “How do we know how smart AI systems are?”

📚 Additional Resources:

• Complexity Podcast, Episode 2: The Relationship Between Language and Thought. We (regrettably) won’t get to this episode of the Complexity podcast in our class, but this is a good time to listen to it if you have time.
• Only a few videos from 3Blue1Brown’s series on Neural Networks are assigned above, but the entire sequence is worth watching for a non-technical overview: https://www.3blue1brown.com/topics/neural-networks
• Witt, Stephen. The Thinking Machine: Jensen Huang, Nvidia, and the World’s Most Coveted Microchip is worth reading in its entirety, but in particular, chapter 6, “Jellyfish” is an excellent non-technical overview of neural nets
• Similarly, the Crash Course: Artificial Intelligence series provides a high level overview of many current AI principles
• There is also a Crash Course: Computer Science for more about computers themselves
• Finally, if you are already statistically inclined, the final episodes of Crash Course: Statistics also cover elements of machine learning and big data analysis building off of core statistical concepts
• Rumelhart, “The Architecture of the Mind: A Connectionist Approach” Mind Design III
• Churchland and Sejnowski, “The Computational Brain” Mind Design III
• Cowie and Woodard, “The Mind is Not (Just) a System of Modules Shaped (Just) by Natural Selection” Mind Design III
• Vaswani et al, “Attention Is All You Need” (Google paper)

TipTip

• “📖 Read”, “🎧 Listen”, and/or “📺 Watch” items are required content for the day, and should be read/heard/watched before class on that day.
• “🌐 Browse” items should be briefly looked at but do not need to be read deeply unless you want to
• “📚 Additional Resources” do not need to be looked at; they are there to serve, if useful, as further references for your debates, final projects, and general edification later.