CHAI Dublin (Chatbot & AI) Meetup

What if I told you that the biggest barrier to AI success isn't bad algorithms—it's that we've forgotten how to think critically about data?
After watching organizations burn through millions on AI projects that fail not from technical limitations but from basic statistical reasoning errors, the pattern is clear: we're letting machines make the same cognitive mistakes that humans have been making for decades.
This talk reveals the systematic methodology that separates AI winners from AI casualties. Using the battle-tested TRACE-M framework (Track record, Reference class, Assumptions, Constraints, Evidence, Measurement) developed from DARPA forecasting tournaments, I'll show you how to dissect any AI claim in under five minutes. You'll learn why Doug Hubbard's Rule of Five can detect AI failures with just five random samples, and master the bias-interruption techniques that turn statistical blind spots into competitive advantages.
Drawing from real disasters where "94.73% accuracy" masked complete system failure, where $2.8B market projections missed reality by 60%, and where hiring algorithms passed every bias test while systematically discriminating, I'll demonstrate the exact process that would have caught each failure before deployment. Forget prompt engineering—the rebels who can systematically evaluate AI claims are the ones building sustainable businesses on top of artificial intelligence.
Key Takeaways:

• The TRACE-M framework: How to evaluate any AI output in under 5 minutes using DARPA forecasting techniques
• Rule of Five in action: Why 5 strategic samples beat 5,000 random data points for quality assessment
• Bias-interruption protocols: Systematic techniques to counter overconfidence, base rate neglect, and survivorship bias in AI evaluation
• Real metrics: How systematic AI skepticism delivers 25-40% better decision outcomes than intuitive evaluation

Target Audience: IT PROFESSIONALS USING AI: including engineers, developers, analysts and testers, IT Architects, IT Project and Product Managers, and all IT business leaders tired of AI promises that don't deliver—and ready to become the people who know why before everyone else figures it out.