AI Automates High-IQ Tasks

Garry Tan shares his perspective on how AI and large language models are revolutionizing knowledge work, particularly emphasizing how proper implementation can automate high-skilled tasks. He believes this transformation presents both opportunities and challenges for businesses and entrepreneurs.

Key Points:

• AI Capability Level:

  • Can now handle tasks typically requiring 130+ IQ with proper workflow and evaluations
  • Currently capable of approximately "120 IQ level work"
  • Best suited for rote, repeating knowledge work

• Business Impact:

  • Companies can operate with smaller teams (2 pizza teams)
  • Potential to build $100M-$1B businesses with just 20 people
  • Corporate America is unprepared for this transformation
  • Will enable more agile startups to outcompete larger incumbents

• Implementation Strategy:

  • Focus on proper prompting and evaluations (evals)
  • Break down complex tasks into smaller steps
  • Need test-driven development approach
  • Avoid "raw dogging" prompts without proper testing

• Market Opportunity:

  • Target businesses spending heavily on knowledge worker teams
  • Look for companies spending hundreds of thousands on offshore teams
  • Replace traditional call centers and business process outsourcing
  • Potential for hundreds of companies to reach tens to hundreds of millions in revenue within 2 years

• Current Limitations:

  • Need to carefully scope AI tasks
  • Can't handle too much in single context window
  • Must break down complex tasks into manageable steps
  • Requires proper evaluation and testing frameworks

• Future Outlook:

  • Next generation models will be 1-2 orders of magnitude more powerful
  • Cost structures likely to decrease by 10x
  • Early movers who establish brands and moats now will benefit most
  • Traditional business moats (Porter's 5 forces) will still matter