You have 4 summaries left

Freakonomics Radio

554. Can A.I. Take a Joke?

Thu Aug 24 2023
Artificial IntelligenceHumorWritingCreativityFuture of Work


This episode of Freakonomics Radio explores the topic of artificial intelligence (AI) and its implications. The host, Adam Davidson, aims to provide a simple and clear explanation of AI and how it should be used. The episode covers various aspects of AI, including its current state, its impact on humor and writing, and the concerns surrounding its use in creative industries. The future role of humans in relation to AI is also discussed. Overall, the episode offers insights into the potential benefits and challenges of AI.


AI's Impact on Humor

AI can simulate emotions and learn patterns to create humor, but currently struggles to consistently produce funny jokes.

Concerns about AI in Writing

There are concerns that AI could weaken the power of writers and lead to a lack of originality in content creation.

The Role of Humans in the Future

The future role of humans depends on whether there are tasks that only humans can do better than AI.


  1. Introduction
  2. Understanding AI
  3. AI and Humor
  4. AI and Writing
  5. AI and Creativity
  6. The Future of AI
  7. The Role of Humans


00:02 - 07:35

  • Adam Davidson, co-creator and former host of NPR's Planet Money, guest hosts this episode of Freakonomics Radio.
  • The three-part series focuses on how to think about artificial intelligence (AI).
  • The main question is whether AI is just the latest trend or a new kind of thing.
  • Encourages more people to be involved in thinking about how AI should be used.
  • The goal is to provide a simple and clear explanation of AI and how to use it.
  • Extreme predictions about AI's impact on humanity are distracting from the current reality.
  • The current generation of AI is not an existential threat but has its own interesting aspects.
  • Chat GPT, an AI tool, can generate impressive text but often produces weird and offbeat results.
  • Can chat GPT be funny? It's not as good as humans yet, according to Lydia Chilton, a professor of computer science at Columbia University.

Understanding AI

07:10 - 14:29

  • AI software uses numbers to predict the next word in a sequence of words.
  • Computers operate on zeros and ones, but can represent words using numbers.
  • Writing rules for generating jokes often results in unfunny jokes.
  • Good old fashioned AI (GoFi) relied on complex rules to achieve outcomes.
  • GoFi's approach to recognizing faces involved breaking them into features and comparing them to a database.
  • GoFi's rule-based approach didn't work well due to lack of computing power and difficulty in writing down all the rules.
  • Human brains learn by building connections among neurons based on sensory input, not by following a list of rules.
  • Neural networks simulate the brain's neurons and are designed to take in data and form connections.
  • The neural network approach has been around since 1943, but only recently became powerful due to advances in computing power and memory capacity.
  • Neural networks require large amounts of data for training, which became available with the advent of the internet.
  • Dumping personal information on the internet provided neural networks with examples of faces, experiences, and thoughts.

AI and Humor

14:01 - 21:19

  • Neural networks were trained with a trove of information from the internet to understand humor.
  • Lydia Chilton, a computer scientist, explored using AI to create humor and understand people better.
  • AI can simulate emotions and learn patterns to create humor.
  • Creativity in humor relies on violating expectations in a structured way.
  • Chilton and her team developed a series of steps for AI to make jokes based on data from The Onion's American Voices section.
  • The AI-generated punchlines showed an understanding of the joke structure but lacked in being funny or making sense consistently.
  • Michael Schur, a TV writer and producer, has reservations about contributing to the advancement of AI language models like GPT.

AI and Writing

21:05 - 28:19

  • Michael Schur is concerned about how AI will be used by industry to weaken the power of writers.
  • The Writers Guild of America (WGA) is striking against the use of AI by studio executives to supplant writers as creators of new ideas for movies and TV shows.
  • The fear is that if AI generates the original idea, writers become hired guns and more rewards go to the studio.
  • The WGA's fight is for the concept of writing as a viable career, as it has become increasingly difficult for young writers to sustain a career in LA or New York.
  • If writing becomes unsustainable, many great stories and talented individuals will be lost.
  • Michael Schur believes that even if AI can only predict and imitate human ideas, there is a danger that people won't be able to tell the difference between AI-generated content and human-created content.
  • Joshua Gans, a professor of strategic management at the University of Toronto, provides a calm perspective on understanding AI without panic or excitement.
  • Gans has written books on AI and runs a program on AI at the National Bureau of Economic Research.

AI and Creativity

27:51 - 35:27

  • AI reduces the cost of prediction by turning information into what is needed.
  • Artificial intelligence is not actively thinking, it uses math to predict.
  • Writing can be decomposed into prompt, writing, and sign off.
  • AI can help improve everyday communication that lacks clarity.
  • AI's ability to imitate and predict based on existing knowledge raises concerns.
  • If AI-generated content dominates, there may be less room for originality and breakthroughs in art.

The Future of AI

35:03 - 42:24

  • The world will be cluttered with garbage and direct and the slurry of other shows and movies.
  • There won't be any room for the good stuff.
  • AI-driven comedy would likely be middle of the road.
  • Human-written comedy can offer a new and unique experience.
  • Creativity is important for economic growth.
  • Economists struggle to define and measure creativity.
  • Property rights, population density, and interest rates are factors economists consider in relation to creativity.
  • Online logo design competitions provide insight into how competition affects creative production.
  • Designers tend to iterate on highly rated designs rather than trying out different ideas.
  • Competition increases creativity up to a certain point, but too much competition discourages effort.

The Role of Humans

42:07 - 48:00

  • Competition in the middle ground leads to creativity and novel work.
  • AI can produce a lot of work in the creative middle ground.
  • When the contest gets crowded, everyone stops competing.
  • The question is whether AI can generate new ideas and forms of writing.
  • Computers and humans both rely on examples to improve their craft.
  • Computers are guessing based on data, but they don't have complete understanding like humans do.
  • The future role of humans depends on whether there are things that only humans can do better than AI.
  • GPT-4 is roughly the size of a squirrel's brain and raises questions about the affordability of training the human brain with AI.
  • Part two will answer big questions about AI and its impact on jobs.