Friends of CST, Full Episodes, Season 1, TechTechTech

Episode 08: The One About Deep Learning

By -c

Unmask the myths and powers of machine learning, deep learning, and AI, the invisible forces shaping our every click and swipe. Watch this episode for the 5 things you need to know about deep learning.

We’re joined by James Kobielus. James is an expert on AI, Data Science, Deep Learning, and Application Development. He was a Big Data Evangelist at IBM and Senior Analyst at Forrester, Burton and others.

We Discuss:

  • What is Deep Learning?
  • Deep Learning’s relationship with Machine Learning, AI, Big Data and Robotics
  • Is it all just more vendor/consultant hype?
  • What are the practical applications of it right now?
  • How do enterprises leverage Deep Learning? What do they need skill wise?
  • Where is it going? How will it be used in the future?

Key Highlights:

  • Machine learning and deep learning are advanced techniques beyond basic machine learning for automating the analysis of data patterns (2:02)
  • There is a lot of hype and misunderstanding around terms like deep learning, big data, and AI (1:38)
  • Deep learning uses neural networks loosely based on how human brains process information (3:03)
  • Big data from social media drove interest in AI techniques like MapReduce in the late 2000s (6:36)
  • AI models need continuous retraining on fresh data or they will fail (17:12)
  • Most SMBs won’t build their own AI but rely on embedded capabilities from vendors (28:07)
  • Applications will increasingly have AI embedded without developers needing data science expertise (32:09)
  • Browser-based machine learning frameworks allow AI in web apps without needing much computing power (32:19)
  • AI hype will die down as people realize the technology requires hard work and doesn’t just run automatically (24:23)

5 Takeaways:

  1. Machine learning and deep learning are advanced techniques that automate the analysis of patterns in data (2:02).
  2. There is a lot of hype and misunderstanding around AI-related terms like deep learning and big data (1:38).
  3. AI models need continuous retraining on fresh data or their performance will degrade rapidly (17:12).
  4. Most small and medium sized businesses will rely on AI capabilities embedded in vendor software rather than building their own (28:07).
  5. Applications will increasingly have AI capabilities embedded without requiring developers to have data science expertise (32:09)

James writes routinely at various outlets:

Check out the Audio-only version of this episode!
Episode 08 – Audio Only