All models are wrong, some are useful…but do we know which ones? – Shaun McGirr, Dataiku


Most organisations have only scratched the surface of the potential value of AI, and we data people often blame corporate culture: “my stakeholders just don’t get it, the inertia is too powerful, people won’t give up Excel, etc…” But practitioners must reflect on how our behaviour contributes: if we treat models as pets to be carefully raised over months or years, is it any wonder data people rarely rise to lead whole organisations? And are our ways of working, maximising lines of code written and bleeding edges surfed, truly cut out for the monumental task of transforming organisations? This session will challenge how we traditionally select which models to build, and offer new ideas on how to become more valuable to your organisation.

Key Takeaways

  • Where did we receive our wisdom on which models to build?
  • How does this get data scientists “trapped” in local optima?
  • Why do we avoid time dimensions in our search for value?
  • What principles will lead us to better choices in what to build? (Hint: we must first recognise the true competition for AI)

Add comment