Switching perspectives
- Challenge: How to get from data to decisions that cause desired outcomes?
- Old perspective: “Fact-based decisions” => Observe pattern (facts), take action
- New perspective: “Outcome-based decisions”
- Current facts alone often do not imply best decisions!
- How do decisions change outcome probabilities?
- “Cause desired outcomes” = make them more likely
- Pattern recognition is not enough
- Must understand goals, cause-and-effect
Our vision for AI technology
Advisory systems with real-world knowledge
- AI products advising human planners, auditors, and leaders
- Help identify better choices
- Out-of-box thinking via AI
- Automated vigilance
- AI proactively monitors and notices relevant changes in real world, suggests response options
- One-click analytics/ML/AI platform
- Causal analytics toolkit (CAT), platform (CAP)
Easy to use
- AI advisor as a smart partner
- Built-in common-sense understanding of likely user goals, constraints, risks, causality
- “Shared intentionality”
- Expert knowledge from web
- Like Watson, but for real => More Holmes, less Watson: Combines deep and wide knowledge with shrewd inference
- Embedded quantitative analytics
- Descriptive, predictive, causal, diagnostic, prescriptive, evaluation, learning, collaboration
- Embedded planning, learning, optimization
- Built-in common-sense understanding of likely user goals, constraints, risks, causality