AI for IA’s: Machine Learning Demystified
I had the pleasure of presenting “AI for IA’s: Machine Learning Demystified” at the 2017 Information Architecture Summit last week in Vancouver, BC, Canada.
I provided the basics of artificial intelligence (AI) and how machine learning works as well as how it is relevant to information architecture (IA).
When machines exhibit intelligence, perceive their environment and take actions to maximize their chance of success at a goal they have artificial intelligence (AI). AI is achieved through the process of training the machine on a specific subject. Training involves spending time transferring human concepts and relationships through documents that are representative of the knowledge we want the machine to know.
The presentation includes discussions on:
- How training is conducted: the general work cycle and a comparison of training methods.
- Watson Knowledge Studio — an example of supervised Machine Learning
- Process and dependence on experts
- Relative importance of knowledge and accuracy
- Goal of artificial intelligence to save time and augment human effort.
- Results of machine learning are more highly trained specialists without additional training. One physician can train a system and that system can then support as many medical staff as needed.
Examples of AI and Cognitive Computing
- 1997 Chess, IBM
- 2016 Go, Google
Self Driving (autonomous) vehicles
Automating labeling of birdsongs
Image Recognition — Google Photos
Automating Repetitive Work — Watson examples:
- Radiology imaging
- Retina imaging
- Customer service representative support
Staples Easy App
Ethics in AI
How might we(1)…
• build systems that have ethical and moral foundation?’
• that are transparent to users?
• teach mercy and justice of law?
• extend and advance healthcare?
• increase safety in dangerous work?
So that we trust machines just as much as a well-trained human?
AI will “extend human experience in profound ways” — Grady Booch, Scientist, philosopher, IBM’er
Explore the technology
Pair with folks to try AI out for yourself — IBM Watson Developer Tools (free trials on most tools)
1) Heavily referenced in this presentation is Grady Booch’s inspiring Ted Talk.