427 AI lecture 1 & 2 quiz
4 views of AI
-Think like a human -Act like a human -Think Rationally -Act Rationally
typical unsupervised ML
-clustering -matrix factorization -graphical model
Key Ingredients in machine learning
-data: collected from past observation -modeling: devised to capture patterns in the data -prediction: apply model to forecast what's going on
classification algorithms
-decision tree -naive bayes -neural networks -support vector machine -logistic regression
regression alogirthms
-linear regression -regression tree -SVM regression
AI history
1943 boolean circuit model of brain mculloch and pitts 1950 Turing "computing machinery and intelligence" imitation game 1956 birth of Ai Dartmouth meeting w/ John , minks & Claude shannon 1955-65 create enthisum (newell & Simon): general prob solver gelertner: geometry theorem prover McCarthy: invention of LISP 1966-73 reality dawns-> realize many AI problems are intractable 1969-85-> success of rule-based expert systems & dev knowledge-based sys 1986-rise of machine learning & neural nets return to popularity 1990 role of uncertainty bayesian networks as knowledge rep framework 1996 AI as science-> integration of learning, reasoning, knowledge rep etc 2001-present resurgence of neural net & dev learning
machine learning
AI technique allows computer to learn without being explicitly programmed
Acting Rationally
Computational intelligence is the study of the design of intelligent agents. -use of decision theory/economics -set of possible actions an agent can take
components of AI
Engineering cognitive science philosophy
big Ai milestones
IBM deep blue beat world chess championion 2003-2007 robot driving drape challenge 2006 face recognition software in cons cameras 2010 ibm beat jeopardy 2016 alpha go beat world champion lee
Ways to evaluate AI
Turing Test Loebner Prize
Act like a human
behaviorist approach -don't care about results as long as you exhibit the same behaviors
challenges with think like a human
challenges -reverse engineer how we reason but thats hard -brain hardware is diff from computer program
Who beat the Turing test?
chatbot named Eugene goostman -calaimed to be 13 year old from Ukraine made by Vladimir and eugeun convinced 33% of judgeds that it was human
typical supervise ML algorithms
classification regression
Think like a human
cognitive science approach
Loebner prize
formal instantiation of the Turing test, held annually with a 100,000 acash prize
utility
gain to an agent for each action/state pair
how can knowledge bottleneck be solved?
in AI may be solved by learning methods instead of hand coded knowledge engineering
turing test
proposed by Turing in 1950 as way of dealing w/ the question whether machines think
problems with thinking rationally
rational agent is predictable & determinstic Deons't act for uncertainty
Thinking Rationally
represent facts about the world via logic -use logical inference as a basis for reasoning about these facts can be useful approach to AI
intractable
stubborn; obstinate; hard to move forward
Supervised vs. Unsupervised learning
supervised ML w/ training examples vs. unsupervised ML: raw data w/o ground truth
