Machine Learning Sucks | Dr. Pieter Buteneers | ML Conference 2018 Spring

Ai

Discover the limitations and challenges of machine learning with Dr. Pieter Butler, discussing the importance of value-driven solutions, limited dataset issues, and more.

Key takeaways
  • Machine learning can replace many jobs, but may not be good at doing something new that it has not seen before.
  • Many startups in AI are just trying to make a quick buck and do not have a business question or solution.
  • Data is expensive and loses value over time, making it hard to keep.
  • Companies should focus on creating value and not just reducing costs.
  • Machine learning is not good at recognizing patterns or making decisions when it is trained on a limited dataset.
  • Overfitting is a major issue in machine learning and can be avoided by keeping it simple.
  • Labeled data is expensive and hard to come by, making it important to create artificial data to learn from.
  • Self-driving cars may not be as good at recognizing patterns as humans.
  • Machine learning can solve many problems, but it is not always the solution.
  • Value-driven economy: focus on creating value and not just reducing costs.
  • AI may not be good at making creative decisions.
  • Many jobs will remain, but new ones will emerge.
  • Machine learning engineers will be in high demand.