Transcript Clouds of Things: Legal Considerations

Microsoft Cloud Computing Research Centre
3rd Annual Symposium, 8th September 2016
Machine Learning, Artificial Intelligence,
and Human Data Interaction
Jon Crowcroft
[email protected]
Artificial Intelligence (AI) / Machine Learning (ML)
 Old story of strong AI v. weak AI
 Most recent headline successes are weak AI
 NLP, ML, Stats at scale
 PCA & Clustering
 Pattern recognition
 Model inferencing….
AI & Human
 There’s a lot of difference between ML & strong AI
 Replacing human intelligence is a ways off according to many experts
 Despite the recent hyped success in Go, still just an NN
 (and recall, chess players really aren’t strong AIs)
 Strong AI gives rise to concern about existential threat
 c.f. Dune (1965), Colossus (1966) or Terminator (1984)
 But we don’t have to go that far to foresee problems…
ML & Human
 But ML already gives us pause for thought about how it interacts
with human domain
 Just as there was a transition from early human computors…
 ....to digital computation
 Also human, accountants, auditors, quants, actuaries, mad men...
 …to analytics
 And this has widespread societal consequences
Human Data Interaction (HDI)
 Agency
 Legibility
 Negotiability
 Present in GDPR…
 For more, see:
 ‘Human-Data Interaction: The Human Face of the Data-Driven Society’
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2508051
Conclusions
 It’s a minefield
 Concentrating on ML today, however we will have a brief discussion
later on AI/HDI
 “The future is already here, it is just unevenly distributed”,
William Gibson (The Economist, 2001)
 So “The proper study of ethics, law & ML is no more nor less than the
proper study of humans, ethics and law” – me, just now
Microsoft Cloud Computing Research Centre
3rd Annual Symposium, 8th September 2016
Machine Learning, Artificial Intelligence,
and Human Data Interaction
Jon Crowcroft
[email protected]