N I V E U R S E I H T T Y O H F G R E U D I B N AI Present and Future Alan Smaill University of Edinburgh, School of Informatics A.Smaill@ed.ac.uk 15/01/19 Alan Smaill AI Present and Future 15/01/19 1/19
AI: Present and Future N I V E U R S E I H T T Y O H F G R E U D I B N Organisation Lecture slots as timetabled. Standard exam at end of semester: exam counts for 75%, coursework for 25%. Alan Smaill AI Present and Future 15/01/19 2/19
AI: Present and Future N I V E U R S E I H T T Y O H F G R E U D I B N Organisation Lecture slots as timetabled. Standard exam at end of semester: exam counts for 75%, coursework for 25%. Formative exercises: there will be a number of unassessed exercises for which the labs are available to drop in and ask questions about. Coursework: there will be one piece of coursework for which there will be a lecture on February 26th, with a deadline on March 15th at 4pm. Alan Smaill AI Present and Future 15/01/19 2/19
Course Info N I V E U R S E I H T T Y O H F G R E U D I B N Course information will be updated on-line as the course proceeds (slides, coursework, references etc). Official description of course: http://www.inf.ed.ac.uk/teaching/courses/aipf/ There is a Piazza site for the course at https://piazza.com/ed.ac.uk/spring2019/infr11180 and all students should register There will be drop-in lab sessions starting in week 3. Alan Smaill AI Present and Future 15/01/19 3/19
Course Info N I V E U R S E I H T T Y O H F G R E U D I B N Course information will be updated on-line as the course proceeds (slides, coursework, references etc). Official description of course: http://www.inf.ed.ac.uk/teaching/courses/aipf/ There is a Piazza site for the course at https://piazza.com/ed.ac.uk/spring2019/infr11180 and all students should register There will be drop-in lab sessions starting in week 3. You cannot take this course if you have taken Informatics 2d! Alan Smaill AI Present and Future 15/01/19 3/19
Course texts N I V E U R S E I H T T Y O H F G R E U D I B N Various recommended reading will appear on the course. 3 texts are central for the course: Russell and Norvig: “Artificial Intelligence: a Modern Approach”, 3rd edition, Prentice Hall, 2016 http://aima.cs.berkeley.edu/ Poole and Mackworth: “Artificial Intelligence”, Cambridge University Press, 2017 https://artint.info/ Blackburn, Bos and Streignitz: “Learn Prolog Now”, College Publications, 2006 http://www.learnprolognow.org/ Alan Smaill AI Present and Future 15/01/19 4/19
Background knowledge N I V E U R S E I H T T Y O H F G R E U D I B N 1. Experience with logic (predicate calculus) will be helpful. 2. A background in probability theory is advisable: Discrete and continuous univariate random variables; Expectation, variance; Joint and conditional distributions. Alan Smaill AI Present and Future 15/01/19 5/19
The course N I V E U R S E I H T T Y O H F G R E U D I B N The course is NOT intended to provide specialist knowledge in parts of AI taught elsewhere in Informatics (Machine Learning, Natural Language Processing, Robotics, Vision, Automated Reasoning, . . . ). In the course, you will be expected to bring your understanding of some specialist areas when discussing questions of how the current different approaches to AI relate to each other, and also what opportunities and dangers there might be in deployment of AI systems in the future. Alan Smaill AI Present and Future 15/01/19 6/19
Topics N I V E U R S E I H T T Y O H F G R E U D I B N Not necessarily in this order: Reasoning agents Logic and inference via Logic Programming Linked data, semantic net and internet search Monte Carlo Tree Search Planning under uncertainty Adversarial search, game playing Probabilistic inference Inductive Logic Programming Approaches to machine learning AI prospects and dangers Ethical and Philosophical issues. Alan Smaill AI Present and Future 15/01/19 7/19
A running theme N I V E U R S E I H T T Y O H F G R E U D I B N In the course, we will often come back to the following questions: What are the relationships between reasoning, computation and prediction in particular AI applications? Alan Smaill AI Present and Future 15/01/19 8/19
A running theme N I V E U R S E I H T T Y O H F G R E U D I B N In the course, we will often come back to the following questions: What are the relationships between reasoning, computation and prediction in particular AI applications? How can we compare symbolic AI systems with subsymbolic and probabilistic systems? We will explain the distinctions involved here as we go through the course. Alan Smaill AI Present and Future 15/01/19 8/19
Some ancient history N I V E U R S E I H T T Y O H F G R E U D I B N Alan Turing’s famous paper from 1950 “Computing Machinery and Intelligence” can be found in many places, eg http://www.abelard.org/turpap/turpap.htm He proposed to replace the question “Can a machine think?” with one where there is a clear way to decide what the outcome is: Can we distinguish between the behaviour of a human and a machine? Alan Smaill AI Present and Future 15/01/19 9/19
Some ancient history N I V E U R S E I H T T Y O H F G R E U D I B N Alan Turing’s famous paper from 1950 “Computing Machinery and Intelligence” can be found in many places, eg http://www.abelard.org/turpap/turpap.htm He proposed to replace the question “Can a machine think?” with one where there is a clear way to decide what the outcome is: Can we distinguish between the behaviour of a human and a machine? It is proposed that a machine may be deemed intelligent, if it can act in such a manner that a human cannot distinguish the machine from another human merely by asking questions via a mechanical link. Turing, 1950 The paper also sketches the capabilities that he believed could be achieved in different areas (vision, natural language, learning, . . . ) Alan Smaill AI Present and Future 15/01/19 9/19
Dartmouth workshop N I V E U R S E I H T T Y O H F G R E U D I B N A proposal to work on “Artificial Intelligence”: The Dartmouth Summer Research project on Artificial Intelligence (1956). We propose that a 2 month, 10 man study of artificial intelligence be carried out during the summer of 1956 at Dartmouth College in Hanover, New Hampshire. The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves. J. McCarthy et al.; Aug. 31, 1955 Alan Smaill AI Present and Future 15/01/19 10/19
Some current AI systems N I V E U R S E I H T T Y O H F G R E U D I B N We’ll look very briefly at some current AI work, to get an idea of the the current state of affairs. Alpha Go Driverless cars Machine translation Alan Smaill AI Present and Future 15/01/19 11/19
Alpha Go N I V E U R S E I H T T Y O H F G R E U D I B N See https://deepmind.com/research/alphago/ From that source: AlphaGo is the first computer program to defeat a professional human Go player, the first program to defeat a Go world champion, and arguably the strongest Go player in history. AlphaGo’s first formal match was against the reigning 3-times European Champion, Mr Fan Hui. in October 2015. Its 5-0 win was the first ever against a Go professional, and the results were published in full technical detail in the international journal, Nature. Alan Smaill AI Present and Future 15/01/19 12/19
Why is this a breakthrough? N I V E U R S E I H T T Y O H F G R E U D I B N Game-playing systems go back to the start of AI (Turing worked on algorithms for playing chess in early 1940s). A time line (where computers better than best human): Checkers: 1994 Chess: 1997 Go: 2015/16 Alan Smaill AI Present and Future 15/01/19 13/19
Techniques N I V E U R S E I H T T Y O H F G R E U D I B N These games are in increasing order of difficulty, according to some analyses of the search spaces involved. Different techniques have been fashionable at different times: fancy heuristics or brute strength? Alan Smaill AI Present and Future 15/01/19 14/19
Techniques N I V E U R S E I H T T Y O H F G R E U D I B N These games are in increasing order of difficulty, according to some analyses of the search spaces involved. Different techniques have been fashionable at different times: fancy heuristics or brute strength? The relentless increase in computational power over the years has also changed the context. Alan Smaill AI Present and Future 15/01/19 14/19
Recommend
More recommend
