どうもありがとうミスターロボート
* I don't speak Japanese. This was copy/pasted from Wikipedia
** This is a talk about Cognitive Computing , not robots or Styx
Slides available online:
http://snugug.github.io/mr-roboto
Before we get started, slides available here
Slide link available again at the end, so no worries
See that Snugug? That's how you can find me online. GitHub and Twitter especially
What is Cognitive Computing?
A new field of computing , for working on Human-complex problems , using machine learning and artificial intelligence , with Human-natural input and output.
The types of problems that can be solved using CC are unlike any problems that have come before and the tech is just now coming out of research
Problems that previously couldn't be worked on by computers due to the input's data or the type of non-linear understanding needed in order to reason about about the domain
AI, machine, and deep learning provide the groundwork to be able to train computers to reason about these Human-complex domains
Beyond keyword text search. Speech, full-sentence and long-form unstructured text, images, audio, video, all become potential inputs and outputs allowing us to interact more like we would with another human
Yah, Yah, Yah. What Does That Mean?
You've already used cognitive apps and didn't know it
Google Now, Siri, both are cognitive applications
Specifically, a common pattern called Personal Assistants
Natural language in
Classify natural language in to core concepts using Natural Language Processing
Parse unstructured documents
Rank solutions based on confidence that solution matches question
Return answer to user
Have the building blocks today: Natural Language Classifier, Natural Language Processing, and Retrieve and Rank
So, Some Examples Then
Let's Play
Three examples
First, a simple application showing what just one cognitive API can do
Next, a more complex one, showing what can be accomplished by combining multiple cognitive APIs in to one application
Finally, what's possible today, not quite by API yet, but what the future holds
All of these examples are built on Watson, with the APIs all having free usage periods and the final application being free to use
Be Kind
Tone Analyzer
Chrome Extension
Watches GitHub Comments
Switches comment submit from green to red if negative
Disables comment submit if negative
Babel Fish
More Advanced, 3 APIs
Speech to Text, Language Translation, Text to Speech
Speak in to computer or phone, have it speak back the translated text
Be able to have live conversations in foreign languages!
As more languages are added, automatically grows in capabilities; no additional programing required
Live Demo
Chef Watson
Do-able today, but not generally available yet
Discovery Advisor, makes connections that humans otherwise couldn't see
Bon Appetite mag and chemical makeup of each ingredient
Give Chef an ingredient, maybe a dish and a style, and it will invent a new recipe for you!
Ingredients generally paired by the regions they grow in; just so happens that food that grows naturally together tastes good together
Because Chef sees beyond this, can find pairings that wouldn't seem obvious at first, like apples and olives
Use it as a base to start cooking from!
Live demo
👍
http://snugug.github.io/mr-roboto
どうもありがとうミスターロボート
* I don't speak Japanese. This was copy/pasted from Wikipedia
** This is a talk about Cognitive Computing, not robots or Styx