005 - Billy Big Models
The Good Stuff
Collaborative AI: Mimicking Human Teams
This chapter explores the multi-agent architecture and its parallels to human collaboration in achieving common goals. It emphasizes the distribution of tasks among agents, questioning the need for larger models while advocating for efficient, AI-driven processes to enhance automation.
The Good Stuff, with Pete and Andy - Episode 5: Billy Big Models
Hosts: Pete and Andy (recording at City Beach, Perth)
Episode Overview: Pete and Andy challenge the belief that more advanced AI models are needed for business transformation, arguing existing models are already sufficient to achieve dramatic cost reductions. They explore frameworks for AI implementation and why businesses should act now rather than waiting for future developments.
Agriculture Mechanization as an AI Parallel (03:40-06:42)
- Before mechanization: 10 farm workers needed to feed 1 person; after: 1 farmer feeds 100+
- Similar dramatic workforce changes could occur in knowledge sectors with AI
The "Billy Big Models" Concept (11:06-15:10)
- Central question: Do we need more advanced AI models for business transformation?
- Pete argues we don't need models better than what existed 9-12 months ago
- Challenging the industry narrative focused on larger models and AGI
Augmentation vs. Agentic Approaches (18:12-22:00)
- Augmentation: Using AI tools to enhance existing human jobs (current focus)
- Agentic: Completely rewiring business processes on an "AI native" standard
- The distinction between "assist" vs. "automate" approaches
Task Decomposition as the Key (23:22-25:55)
- Breaking down jobs into "subatomic" tasks is crucial
- Each agent only needs to do one small task well, not an entire job
- 100 simple agents is exponentially easier than creating one complex agent
- Human tasks ($100) → LLM tasks ($10) → silicon/code tasks ($1)
Traditional Business Model Problems (29:10-32:38)
- Companies adding AI as "tool creep" rather than rethinking from first principles
- Pursuing 5-15% efficiency gains when competitors might achieve 90% cost reduction
- Processes designed around humans remain expensive even with AI tools
E-commerce Parallel to AI Transformation (35:44-39:12)
- Limited approach: existing stores simply adding websites (e.g., Myer)
- Transformative approach: complete rethinking (e.g., Warby Parker)
- Question of whether incumbents will be displaced by AI-native businesses
Obstacles to Business Transformation (43:56-49:57)
- Lack of understanding about AI implementation approaches
- Insufficient intent, time, and money to transform operations
- Middle management's disincentives to reduce team size
- Waiting for "AGI" as an excuse for inaction
The "Learn to Surf" Analogy (53:28-55:51)
- Business transformation requires reading patterns like a surfer
- Technology shifts as waves that businesses must learn to ride
- The ocean/wave is more powerful than you - you must adapt to it
Agency as the New Competitive Advantage (56:46-01:00:02)
- High-agency individuals will benefit most in the next 5-10 years
- Agency may become "the new IQ" for harvesting opportunity
- The need for adaptability rather than rigid expertise
Advancements Beyond Bigger Models (01:03:40-01:10:45)
- Recent key advancements weren't bigger models but new techniques:
- Multi-agent architectures mirroring how teams of people approach problems.
Notable Quotes:
- "If you're sat around waiting for AGI and AGI takes 10 years, and right now I can take 80% of the cost out of a business line, I will crush you in the market."
- "What would this process look like if I now had unlimited access to intelligence?"
- "Agency is going to be the new IQ. That's the standard for harvesting opportunity."
The AI-powered Podcast Player
