In this debut episode, I sit down with my buddy Marcel Gregoriadis — a PhD candidate in Computer Science — to talk about everything AI.
- We break down exactly how LLMs work and the most common mistakes we all make using AI.
- We cover exciting research out of Harvard and Wharton showing how working with AI can boost productivity by 40% and how to harness AI’s creativity for your projects.
- But there is the danger of using AI on autopilot and as a shortcut.
- An MIT Media Lab study found that 83% of participants couldn’t remember their own work and that outsourcing work to AI dramatically reduced brain activity and connectivity.
- We also share 3 practice-backed protocols to get smarter, not dumber with AI.
- Finally, we speculate on what this all means for our generation entering the job market.
Timestamps:
00:00 Intro to Balance & Breakthrough
00:20 Teaser and Take-Aways
1:24 Welcome to the Debut Episode: Meet my buddy Marcel
01:27 AI Fundamentals: Breaking down LLMs
07:44 The Breakthrough: How the attention mechanism changed everything
10:45 How the transformer architecture fixed the "Chinese whispers" context problem
11:27 How LLMs became smart: Multi-head attention
21:28 Mistake #1: Context Rot - Why your chat gets dumber as it grows
23:01 Mistake #2: Don’t just tell chatbots what to do, create the vibe and persona
29:05 The Manager and Minions: How to think about AI collaboration
30:12 Danger of Using AI: Shortcutting hard thinking and hitting brick wall
31:41 The Harvard/Wharton Study: A 40% quality boost for consultants
34:32 The Autopilot Danger: Falling asleep at the AI wheel
35:58 Groupthink: Why AI can make organisations less creative
38:31 The Creativity Test: Why ChatGPT beats MBA students 35 to 5 in start-up competition
42:47 The MIT Brain Study: 83% of users forgot their own work!
48:59 Protocol 1: The "Brain First" Rule: Do the hard thinking first, then bring in AI to augment your intelligence
50:46 Protocol 2: The “Lean Startup” Prompting Method: Using AI iteratively to stop it from building a "nonsense cathedral"
53:04 Protocol 3: The “Sequential Thinking” Protocol: The trick to force the AI to critique its own answers
1:00:28 The Future of Work: Are juniors becoming irrelevant?
1:04:34 Dystopia vs. Utopia: Universal Basic Income and the 10-hour work week?!
1:09:45 The Paradox: Use AI less to use it more
1:10:56 Closing Tradition: Marcel’s search for balance and his breakthroughs
🔗 Sources Mentioned in this Episode:
📚 Books:
- Co-Intelligence: Living and Working with AI by Ethan Mollick: https://www.amazon.com/Co-Intelligence-Living-Working-Ethan-Mollick/dp/059371671X
📑 Research Articles and Studies:
- Google Study: Attention Is All You Need (2017): https://proceedings.neurips.cc/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf
- The Harvard/Wharton/BCG Study: Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality (2023): https://www.hbs.edu/ris/Publication%20Files/24-013_d9b45b68-9e74-42d6-a1c6-c72fb70c7282.pdf
- The MIT Media Lab Study (2025): Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task: https://www.media.mit.edu/projects/your-brain-on-chatgpt/overview/
👨💻 Connect with the Guest
Marcel Gregoriadis (PhD Candidate in Computer Science, TU Delft)
LinkedIn: https://www.linkedin.com/in/marcel-gregoriadis/
🎙️ Connect with the Host:
Noah Maurer (PhD Student in Management and Entrepreneurship, Imperial Business School)
Instagram: https://www.instagram.com/noah_ehw/
LinkedIn: https://www.linkedin.com/in/noah-maurer/