Filler lecture Duolingo interactions OLD AI topic theorem

Filler lecture Duolingo interactions [OLD] AI topic: theorem proving New: Matchmaking

Duolingo interactions • What is automatic? • What is done by duolingo staff? – Prompted by normal activity – Prompted by questions – Prompted by feedback (user corrections) • There was one instance of this. • What is cloud source, that is, members of the community can contribute? – User poses question or comment (insight? ) • Zenon on Seat car

Duolingo • Basic business case for duolingo may be changing • Was: real-world translation • Future: – Web documents – Documents submitted by users – ? ? ? • News: http: //techcrunch. com/2012/09/17/duolingo-raises-15 m-series-bround-lead-by-nea-will-expand-to-more-languages-and-to-mobile-soon/

Theorem proving • Recall comment: effort tends to not be called AI once it works. • Theorem proving techniques used in checking software. Other?

Theorem proving • Need to express everything in formal logic. Various methods. • Assume the negative of what you want to prove. • Try everything with everything else…. • If you reach an impossibility: A and NOT A, then the original proposition has been proven.

Example • http: //wwwrci. rutgers. edu/~cfs/305_html/Deduction/Res olution. TP. html

News • Economics Nobel Memorial prize awarded to Shapley and Roth for work on matchmaking. • This included algorithms AND experiments. • http: //cse 331. wordpress. com/2009/09/11/lec t-4 -gale-shapley-algorithm/

Preview • After midterm, we will work on computability (including what came to be called Turing machines) and complexity (how much resources taken by an algorithm). • Measure of time (or space) given using something called big oh notation. • Claim that Gale-Shapley is O(n 2).