Semantic Image Search: a non-technical introduction to machine learning

In which I give a non-technical introduction to machine learning in the context of Google's recent announcements about using deep learning techniques for semantic image search, explain why you should use it and discuss some of the ethical challenges it raises.

Transcript

1. SEMANTIC IMAGE SEARCH Ade Oshineye
2. www.oshineye.com/+
3. www.oshineye.com/+
4. WHAT IS GOOGLE?
5. THE MISSION & THE TOOLS “Google’s mission is to organize the world’s information and make it universally accessible and useful.”
6. MACHINE LEARNING Learning by example not learning by programming Here are some features Classify and generalise
7. MACHINE LEARNING: TERMINOLOGY 1 Features 2 Training 3 Clustering 4 Classification
8. MACHINE LEARNING: TERMINOLOGY 1 Supervised learning 2 Unsupervised learning 3 Semi-supervised learning 4 Deep learning
9. HOW WOULD YOU BUILD IMAGE SEARCH?
10. SEARCHING FOR RED PHOTOS
11. SEARCHING FOR GREEN PHOTOS
12. SEARCHING FOR YELLOW PHOTOS
13. CONVENTIONALLY 1 Identify feature 2 Write feature extractor
14. SEARCHING FOR BORIS BIKES Californians wouldn’t know about these
15. SEARCHING FOR PARIS & OTHER CITIES 3400 cities with 100k population
16. !SCALE The world’s information...
17. DEEP LEARNING: PROBLEM STATEMENT 1 Too many features 2 Too many labels 3 Too much data
18. DEEP LEARNING: SOLUTION 1 Unsupervised (mostly) 2 Machine learning for feature extractor 3 Hierarchy of features and pre-training 4 Machine learning for classifier Based on applying academic work at unprecedented scale
19. 2K classes from Freebase 5K images per class for training
20. GENERALISE Infer structure and patterns
21. Skipping over a ton of other work by other teams and systems to turn this into a useful production service
22. CHALLENGES 1 Feedback 2 Finding applications 3 The Highlander test and face recognition 4 The screwdriver test and ethics
23. ACCESSIBLE AND USEFUL “We systematically overestimate the value of access to information and underestimate the value of access to each other” Clay Shirky
24. THE FUTURE IS UNWRITTEN
25. THANK YOU