The Art of Manliness Podcast

Episode: Co-Intelligence — Using AI to Think Better, Create More, and Live Smarter
Date: April 15, 2025
Host: Brett McKay
Guest: Ethan Mollick, Professor at Wharton Business School and author of Co-Intelligence: Living and Working with AI

Episode Overview

This episode explores the revolutionary impact of artificial intelligence, particularly large language models (LLMs) like ChatGPT, Claude, and Gemini, on daily life, work, education, and creativity. Host Brett McKay interviews Ethan Mollick, a leading researcher, professor, and author, about how AI is rapidly becoming a co-intelligent partner, reshaping our expectations of productivity and forcing us to re-examine what it means to be human in an age when machines can mimic, and sometimes surpass, human skills.

Mollick shares actionable advice for using AI tools thoughtfully—discussing how anyone can harness LLMs to free up creativity, improve learning, and enhance productivity, without yielding their agency or becoming overwhelmed by existential dread.

Key Discussion Points and Insights

1. The True Nature of AI and LLMs

[02:55–05:52]

• Definitions:

  • Early AI focused on "machine learning"—computers recognizing patterns in data and making predictions (e.g., weather forecasts, movie recommendations).
  • Large language models like ChatGPT have expanded AI’s reach into human language, enabling much more nuanced and creative applications.
  • These models work by ingesting huge datasets (everything from Wikipedia to Enron emails to fan fiction) and "learning" statistical relationships between words ("tokens").

• Mollick’s analogy:

  • “It’s basically like the world’s fanciest autocomplete that happens to be right a lot of the time.” (05:48)

2. How LLMs Create Images and Multimodal Breakthroughs

[06:00–07:09]

• Early image creation: Used “diffusion” models (turning noise into images, but often lacking intelligence).
• Recent Leap (2025): LLMs like ChatGPT 4.40 and Google Gemini now generate images directly using the same stepwise prediction method as with words—making more accurate, multimodal creations.

3. Comparing LLMs: GPT, Claude, Gemini

[07:09–08:32]

• Models vary, but rapid evolution means most "frontier" models are increasingly similar.
• Advice: Always use the largest, latest model available—capabilities improve fast.

4. AI Surpassing Human Performance on Tests

[08:32–10:19]

• AI now outperforms humans in general knowledge and specialized exams (e.g., GPQA, the Bar, AP exams).

• “The AI will beat most humans in most tests.” – Mollick (10:19)

• New, harder tests (“Humanity’s Last Exam”) are quickly being tackled by AIs as well.

5. The Limits, Failures, and Quirks of LLMs

[12:28–16:24]

• Jagged Frontier: AIs have strange strengths/weaknesses (e.g., writing brilliant creative sonnets, but struggling to count letters in a word…though even these “weaknesses” are disappearing).
• Hallucinations:
  • AIs sometimes make up plausible-sounding but untrue facts or quotes (“hallucinations”).
  • “They make things up that sound real…The fact that hallucinations are right so much of the time is kind of weird.” – Mollick (13:34)
  • Hallucination rates are dropping as models get better—with recent models hitting 0% hallucination on medical question sets (15:14).
• Creativity:
  • LLMs excel at creativity, often besting average humans in idea generation (e.g., business class startup ideas: 35/40 top ideas were from AI, not students; 17:13).

6. Worries and Existential Threats

[17:13–20:47]

• Societal Concerns:

  • Short-term: Fake images, manipulation, job disruption.
  • Long-term: Existential threats (machine superintelligence pursuing its own goals, “paperclip problem,” loss of control over essential systems).

• Mollick’s Sleepless Nights Theory:

  • “If you haven’t had the sleepless nights while using AI, it’s because you haven’t used it enough or gone deep with it…You start to worry: what does this mean for my job? For my kids’ jobs?” – McKay (20:47)

7. AI as Co-Intelligence: Guidelines for Effective Use

[22:00–31:25]

Mollick outlines four core guidelines for using AI well:

1. Always Invite AI to the Table

• Use AI in your domain of expertise for at least 10 hours.
• “Just use it. If you haven’t put 10 hours in because you’re avoiding it for some reason, you just need to do it.” (23:26)

2. Be the Human in the Loop

• Use AI to enhance your own decision-making and agency—not surrender them.
• “You want to integrate AI into your work in a way that increases your own importance and control…Use it to support what you do to do it better.” (24:20)

3. Treat AI Like a Person (For Prompts!)

• Not philosophically or emotionally, but as a conversational partner for more productive prompting.
• “If you treat it like a human being, you are 90% of the way there to prompting it…Programmers are sometimes worse with AI because they want to work like software code.” (26:26)
• Don’t forget: AI is ultimately software, not sentient.

4. Assume This Is the Worst AI You Will Ever Use

• The rapid pace of improvement means the only constant is change.
• “If AI can’t do something now, it’s probably worth checking in a month or two to see if it can do it then.” (31:27)

8. Prompt Engineering: Getting the Most out of LLMs

[39:36–41:37]

Mollick’s Four Prompting Techniques:

1. Be Direct: Clear, unambiguous instructions.
2. Provide Context: The more background you give, the better the results (include documents, describe the use-case, or assign a “persona”).
3. Chain of Thought Prompting: Ask for stepwise reasoning—have AI spell out its thinking.
4. Few-Shot Prompting: Give specific examples of what you want (and what you don’t).

• Memorable user example: Brett used an LLM as a simulated expert doctor/therapist, provided a photo and symptoms, and got better rehab advice than from his doctor/therapist (42:41).

9. Real-World Applications: Work and Productivity

[43:38–49:03]

• Workplace tips:

  • Use AI for advice, brainstorming, writing, summarizing, coding, or research.
  • Example: Small businesses in Kenya saw 12–18% profit increases just by asking AI for business advice.
  • Don’t limit yourself to one answer or idea—ask for many, then curate.
  • “Think maximalist: Ask it for 30 options, not one. It doesn’t get tired.” (44:01)
  • Use AI for “80%” of generic tasks; focus your unique abilities on the remaining 20% of complex or creative work (47:23).

• Delegation Ethics:

  • Keep personally meaningful or ethically sensitive tasks for humans (e.g., grading as a professor).
  • Delegate boring or weak areas to AI.

10. Creativity, Humanities, and Gaining an Edge

[37:25–39:36, 49:03–51:14]

• Those with broader general/humanities knowledge get more unique, tailored results—because they can prompt AI with richer connections, references, and styles.
• AI art and writing only reflect prompts—your breadth of imagination matters.
• “You can invoke styles, personas…If you have a wide set of knowledge, you get the AI to work in those styles and create things others can’t.” (37:25)
• In creativity, AI can handle the “gruntwork” and help fill in gaps, but humans remain best at top-level skill or vision (for now).

11. Learning, Education, and the “Homework Apocalypse”

[51:14–58:52]

• Disruption:

  • 70% of students now use AI for homework; “AI detectors” are unreliable.
  • Like calculators in math, assignments must shift—focus on in-class, active learning and oral questioning to assess real understanding.

• Opportunities:

  • Incorporate AI as a learning assistant, not a substitute.
  • Personalized AI tutors (with teacher oversight) can boost learning—studies cited in Harvard, Nigeria showing clear gains.
  • “General knowledge is more important than ever. Expertise is more important than ever.” (54:14)

• For Parents:

  • Use AI to help coach kids, not just hand them the answers—e.g., have it create practice quizzes, explain concepts, or offer alternative explanations.

12. Lifelong Learning with AI

[58:52–60:33]

• LLMs serve as effective "co-learners" or "co-thinkers" for adults as well—Brett uses AI as a literature tutor for deep reading; a quantum physicist uses AI to ask good questions, sparking new ideas.

13. The Future of AI and Final Thoughts

[60:33–62:10]

• Rapid Improvement:
  • New breakthroughs (“reasoners”—models that think before answering) mean AIs might achieve “artificial general intelligence” (AGI) within a few years.
  • Nobody knows precisely where this will lead; being adaptable, learning with the technology, and using it creatively are key.
  • AI is not just “happening to us”—we have agency in shaping its role in work and society.

Notable Quotes & Memorable Moments

• “It’s very hard to use these systems and really use them…I find, by the way, a lot of people kind of bounce off them precisely because they feel like this kind of dread and they sort of walk away.” – Ethan Mollick [19:57]
• “If you haven’t had the sleepless nights while using AI, it’s because you haven’t used it enough or gone deep with it.” – Brett McKay [20:47]
• “Treat AI like a human is a technique for using AI. It is not a philosophy.” – Ethan Mollick [31:21]
• “If AI can’t do something now, it’s probably worth checking in a month or two to see if it can do it then.” – Ethan Mollick [31:34]
• “Think maximalist: Ask it for 30 options, not one. It doesn’t get tired.” – Ethan Mollick [44:01]
• “We still lack thinking partners in the world…So it can help you spur your own thinking.” – Ethan Mollick [59:44]
• “The most intelligent way to use these tools is to supplement your thinking, not replace it.” – Brett McKay [60:33]

Timestamps for Critical Segments

• [02:55] — Defining “artificial intelligence” and LLMs
• [06:00] — How AIs create images; new multimodal capabilities
• [08:41] — AIs surpassing human performance on knowledge tests
• [13:34] — Hallucinations: why AI makes up facts
• [17:28] — Societal and existential worries about AI
• [23:26] — Practical guideline: 10 hours with AI, invite it to your work
• [39:36] — Four practical prompting techniques
• [43:38] — Real-world workplace advice; maximizing AI’s value
• [51:55] — The “homework apocalypse” and how schools/teachers adapt
• [58:52] — Tips for parents using AI as a tutor
• [60:40] — The future: AGI and ongoing rapid change

Summary

This episode offers a balanced, deeply practical look at how anyone—from writers and business owners to parents and teachers—can use modern AI as a "co-intelligent" partner. Rather than a force of pure disruption (or outright replacement), LLMs can help individuals amplify their strengths, automate their weaknesses, and think more expansively. However, the hosts and guest stress: mastery of these tools requires experimentation, discernment, and ongoing learning—as the technology, and its challenges, are evolving faster than anyone anticipated.

Further Resources:

• Ethan Mollick’s newsletter: oneusefulthing.org
• Book: Co-Intelligence
• Prompt list for parents/teachers: [Generative AI Lab at Wharton]

This summary maintains the conversational, insightful tone of the original episode, distilling its best practical advice and existential insights for listeners who aim to think—and live—sharper with AI.