IAI1 Lecture 1 2024 PDF

Lecture 1 - Introduction: What is Artificial Intelligence? Henry Brighton Introduction to Artificial Intelligence 1 (822184) People Lecturers: I dr. Henry Brighton (course coordinator) I dr. Noortje Venhuizen I dr. Emmanuel Keuleers I dr...

Lecture 1 - Introduction: What is Artificial Intelligence? Henry Brighton Introduction to Artificial Intelligence 1 (822184) People Lecturers: I dr. Henry Brighton (course coordinator) I dr. Noortje Venhuizen I dr. Emmanuel Keuleers I drs. Stijn Rotman (practicals coordinator) Teaching Assistants: I Maria Ene I Jan van Gestel 39 Why study artificial intelligence? 3 39 Artificial Intelligence: The big questions 1. Can we build machines that think? 2. Can we build machines that learn? 3. Can we build machines that are more intelligent than us? 4. Can we build machines that are creative? 5. Can we build machines that have emotions? 6. Can we build machines that are conscious? 4 39 What we read in the press... 5 39 Killer robots: Experts warn of “third revolution” in warfare BBC News, August 2017 Killer robots: Experts warn of “third revolution” in warfare I AI experts warn the United Nations of a third revolution in warfare. BBC News, August 2017 Killer robots: Experts warn of “third revolution” in warfare I AI experts warn the United Nations of a third revolution in warfare. I Fully autonomous weapons that engage targets without human intervention. BBC News, August 2017 Killer robots: Experts warn of “third revolution” in warfare I AI experts warn the United Nations of a third revolution in warfare. I Fully autonomous weapons that engage targets without human intervention. “Once developed, they will per- mit armed conflict to be fought at a scale greater than ever, and at timescales faster than humans can comprehend” BBC News, August 2017 How safe can artificial intelligence be? BBC News, September 2015 7 39 How safe can artificial intelligence be? On the one hand: I Stephen Hawking says “AI could spell the end of the human race”. I Elon Musk says it’s “like summoning the demon”. BBC News, September 2015 7 39 How safe can artificial intelligence be? On the one hand: I Stephen Hawking says “AI could spell the end of the human race”. I Elon Musk says it’s “like summoning the demon”. On the other hand: I Brett Kennedy (JPL) says “I have first-hand knowledge of how hard it is for us to make a robot that does much of anything”. I Alan Winfield (U. of Bristol) says “fears of future super intelligence - robots taking over the world - are greatly exaggerated”. BBC News, September 2015 7 39 AI Hype? What about machine learning? AI Hype? What about machine learning? David Harding AI Hype? What about machine learning? Founder and CEO of Winton Capital, a top 10 hedge fund. Using machine learning: I 1980s: Started investing I 1997: $1.6 million in assets I 2000: $150 million I 2004: $1 billion I 2006: $4.8 billion I 2009: $12.4 billion I 2017: $28.5 billion David Harding AI Hype? What about machine learning? Founder and CEO of Winton Capital, a top 10 hedge fund. Using machine learning: I 1980s: Started investing I 1997: $1.6 million in assets I 2000: $150 million I 2004: $1 billion I 2006: $4.8 billion I 2009: $12.4 billion I 2017: $28.5 billion “I think the public debate about AI and machine learning is nine parts hype to one part substance” David Harding (2018) 8 39  Be very, very skeptical Theodore Roszak. (1986). Smart computers at insecure stage. New Scientist, 110, p.47. 9 39  Be very, very skeptical “AI’s record of barefaced public deception is unparalleled in the annals of academic study” Theodore Roszak Theodore Roszak. (1986). Smart computers at insecure stage. New Scientist, 110, p.47. 9 39  Be very, very skeptical “AI’s record of barefaced public deception is unparalleled in the annals of academic study” Theodore Roszak I AI researchers have been making bold claims since the late 1950s. I Repeatedly, AI’s progress has been much slower than expected. I It’s easy to imagine full AI, hence easy to convince/scare people. Theodore Roszak. (1986). Smart computers at insecure stage. New Scientist, 110, p.47. 9 39 “The AI problem is one of the hardest science has ever undertaken” Marvin Minsky (1982) 10 39 An overview of today’s session 1. What is artificial intelligence? 2. What is the goal of artificial intelligence? 3. What does artificial intelligence, as a field of research, look like? 4. This course: What will you learn and how will you be evaluated? 11 39 What is Artificial Intelligence? Artificial intelligence is: 12 39 What is Artificial Intelligence? Artificial intelligence is: I “the science and engineering of making intelligent machines” — John McCarthy. 12 39 What is Artificial Intelligence? Artificial intelligence is: I “the science and engineering of making intelligent machines” — John McCarthy. I “the science of making machines do things that would require intelligence if done by men” — Marvin Minsky (1968) 12 39 What is Artificial Intelligence? Artificial intelligence is: I “the science and engineering of making intelligent machines” — John McCarthy. I “the science of making machines do things that would require intelligence if done by men” — Marvin Minsky (1968) I “The exciting new effort to make computers think... machines with minds, in the full and literal sense” — John Haugeland (1985) 12 39 What is Artificial Intelligence? Artificial intelligence is: I “the science and engineering of making intelligent machines” — John McCarthy. I “the science of making machines do things that would require intelligence if done by men” — Marvin Minsky (1968) I “The exciting new effort to make computers think... machines with minds, in the full and literal sense” — John Haugeland (1985) I “The art of creating machines that perform functions that require intelligence when performed by people” — Ray Kurzweil (1990) Let’s break the issue down... 12 39 What do we mean by “artificial”? 13 39 What do we mean by “artificial”? If machine intelligence is artificial, what is “real” intelligence? 13 39 What do we mean by “artificial”? If machine intelligence is artificial, what is “real” intelligence? I Must “real” intelligence be made from biological stuff? I This is known as carbon/protoplasm chauvinism. I What if we work out how to engineer biological agents? 13 39 What do we mean by “artificial”? If machine intelligence is artificial, what is “real” intelligence? I Must “real” intelligence be made from biological stuff? I This is known as carbon/protoplasm chauvinism. I What if we work out how to engineer biological agents? I Must “real” intelligence be the product of biological evolution? I Should nature be the only source of “real” intelligence? I What if we work out how to evolve biological agents? 13 39 What do we mean by “artificial”? If machine intelligence is artificial, what is “real” intelligence? I Must “real” intelligence be made from biological stuff? I This is known as carbon/protoplasm chauvinism. I What if we work out how to engineer biological agents? I Must “real” intelligence be the product of biological evolution? I Should nature be the only source of “real” intelligence? I What if we work out how to evolve biological agents? So, we seem to mean two things: I By artificial, we might mean non-biological. Here, it’s all about the kind of stuff we use to build an agent. I By artificial, we might mean constructed by humans. Here, it’s all about the origin of the agent, and who designed and built it. What do we mean by “intelligence”? I “Few concepts in psychology have received more devoted attention and few have resisted classification so thoroughly” — A.S. Reber What do we mean by “intelligence”? I “Few concepts in psychology have received more devoted attention and few have resisted classification so thoroughly” — A.S. Reber I “Intelligence is the computational part of the ability to achieve goals in the world. Varying kinds and degrees of intelligence occur in people, many animals and some machines” — John McCarthy What do we mean by “intelligence”? I “Few concepts in psychology have received more devoted attention and few have resisted classification so thoroughly” — A.S. Reber I “Intelligence is the computational part of the ability to achieve goals in the world. Varying kinds and degrees of intelligence occur in people, many animals and some machines” — John McCarthy An Example? Meet Nigel... An Example? Meet Nigel... Nigel is: A Neato Robotics Botvac Connected Nigel is one of the most advanced robot vacuum cleaners on the market: I Laser range finder. I Bump sensors. I Drop sensors. I Two wheels with elevation control. What’s special about Nigel? I Nigel has a charging station he returns to when his battery drains. I Nigel doesn’t move randomly, he forms a plan about how to clean. I Nigel does this by investigating and building a map of the room. I Is Nigel an example of artificial intelligence? What do you think? 16 39 Nigel and the big AI questions 1. Can Nigel think? 2. Can Nigel learn? 3. Is Nigel more intelligent than us? 4. Is Nigel creative? 5. Does Nigel have emotions? 6. Is Nigel conscious? 17 39 An overview of today’s session 1. What is artificial intelligence? 2. What is the goal of artificial intelligence? 3. What does artificial intelligence, as a field of research, look like? 4. This course: What will you learn and how will you be evaluated? 18 39 The Dartmouth Summer Research Project on Artificial Intelligence (1956) 19 39 The Dartmouth Summer Research Project on Artificial Intelligence (1956) The first AI “hypothesis” “Every aspect of learning or any other feature of intelligence can in principle be so precisely defined that a machine can be made to simulate it” 19 39 Artificial intelligence has grown into a field of study with many goals. 20 39 Strong AI vs. Weak AI “Strong” Artificial Intelligence “Weak” Artificial Intelligence The goal is to understand the human vs. The goal is to develop intelligent mind/brain as a computational device. machinery that is not necessarily like us, but nevertheless solves some problem. Specific in the sense that we are vs. General in the sense that any kind of focused on the human mind/brain. problems/techniques might be entertained. General in the sense that humans can vs. Specific in the sense that a single, solve an astonishingly wide range of isolated problem is usually considered. problems. Based on the belief that we can vs. We are interested in machines that construct machinery capable of solve tricky problems in ways perhaps thought, emotions, and any other not found in nature, independent of quality we attribute to humans. whether “strong AI” is possible. 21 39 AI as Science vs. AI as Engineering AI as Science AI as Engineering Reverse-engineer the mind/brain. vs. Engineer clever machines. The object of interest exists in nature. vs. The objects of interest are unknown, and are to be discovered/created. Investigate the hypothesis that vs. Independent of human nature, can we humans are machines. construct clever machines? AI as a science could fail. vs. AI as engineering has already succeeded. 22 39 Artificial General Intelligence vs. Narrow AI Artificial General Intelligence Narrow AI Real progress is likely to come from vs. Real progress is likely to come from viewing intelligent systems as whole. breaking the big problems down, and tackling smaller problems. Set human level performance as the vs. Seek incremental improvements. goal, and abandon efforts incapable of Rome wasn’t buit in a day. reaching this goal. Seek early integration of AI’s vs. Let each subfield solve it’s own subfields. problem first. 23 39 The Rational vs. The Psychological Should the study of AI be guided by the psychological (our understanding of how humans work), or the rational (our understanding of logic, probability theory, etc.)? The Rational vs. The Psychological Should the study of AI be guided by the psychological (our understanding of how humans work), or the rational (our understanding of logic, probability theory, etc.)? This distinction impacts both on the kind of systems we study, and the kind of behavior we desire... Psychological Rational Thinking (Mechanisms) Thinking Humanly: AI guided Thinking Rationally: AI guided by the study of cognitive by the study of rational mechanisms. processes of optimization. Acting (Behavior) Acting Humanly: We want Acting Rationally: We want machines that behave like machines that behave humans. optimally. See: Russell, S. & Norvig, P. (2009). pages 1-5. An overview of today’s session 1. What is artificial intelligence? 2. What is the goal of artificial intelligence? 3. What does artificial intelligence, as a field of research, look like? 4. This course: What will you learn and how will you be evaluated? 25 39 Artificial Intelligence: The Big Questions 1. Can we build machines that think? 2. Can we build machines that learn? 3. Can we build machines that are more intelligent than us? 4. Can we build machines that are creative? 5. Can we build machines that have emotions? 6. Can we build machines that are conscious? 26 39 These questions are not just questions for artificial intelligence. 27 39 Artificial Intelligence and Cognitive Science 28 39 29 39 Most AI research doesn’t tackle the big AI questions directly. 30 39 The Fragmentation of AI Research The big AI questions Robotic agents Software agents Locomotion... Learning Reasoning... Supervised Unsupervised Reinforcement... learning learning learning Connectionist methods... Generative adversarial... networks 31 39 The Fragmentation of AI Research The big AI questions Robotic agents Software agents Locomotion... Learning Reasoning... Supervised Unsupervised Reinforcement... learning learning learning Connectionist methods... Most AI research is Generative disconnected from the adversarial... “big AI questions”. networks 31 39 Artificial Intelligence: Some bad news and some good news... 32 39 Trends in Artificial Intelligence: 1956-2004 Source: Google Ngram Viewer. 33 39 Summary 34 39 Summary 1. What is artificial intelligence? I Artificial: Computational agents, either robotic or software. I Intelligence: Agents capable of achieving goals in the world. 34 39 Summary 1. What is artificial intelligence? I Artificial: Computational agents, either robotic or software. I Intelligence: Agents capable of achieving goals in the world. 2. What is the goal of artificial intelligence? I Strong vs. weak. I Science vs. engineering. I General vs. narrow. I Rational vs. Psychological 34 39 Summary 1. What is artificial intelligence? I Artificial: Computational agents, either robotic or software. I Intelligence: Agents capable of achieving goals in the world. 2. What is the goal of artificial intelligence? I Strong vs. weak. I Science vs. engineering. I General vs. narrow. I Rational vs. Psychological 3. What does artificial intelligence, as a field of research, look like? I AI is closely allied to several disciplines. I AI includes a very wide, arguably fragmented, range of research areas. I AI has had it’s problems! Is is a myth? Can it deliver? 34 39 An overview of today’s session 1. What is artificial intelligence? 2. What is the goal of artificial intelligence? 3. What does artificial intelligence, as a field of research, look like? 4. This course: What will you learn and how will you be evaluated? 35 39 Overview: Introduction to Artificial Intelligence 1 (822184) Date Where When What 1 Monday, 26th August WZ1 16:45 - 18:30 Lecture 1: What is Artificial Intelligence? 2 Tuesday, 3rd September MKZ 1 10:45 - 12:30 Lecture 2: Cognition and Computation Thursday, 5th September Practical 1: Problem Representation 3 Tuesday, 10th September Aula 08:45 - 10:30 Lecture 3: Classical Search Thursday, 12th September Practical 2: Classical Search 4 Tuesday, 17th September Aula 08:45 - 10:30 Discussion 1: Margaret Boden 5 Tuesday, 24th September GZ 101 10:45 - 12:30 Lecture 4: Local Search 6 Tuesday, 1st October GZ 101 08:45 - 10:30 Discussion 2: Turing Thursday, 3rd October Practical 3: Local and Adversarial Search (1) 7 Tuesday, 8th October MKZ 1 14:45 - 16:30 Lecture 5: Adversarial Search TBA Midterm Exam 8 Tuesday, 22nd October MKZ 1 10:45 - 12:30 Lecture 6: Problem Solving Under Uncertainty Thursday, 24th October Practical 4: Local and Adversarial Search (2) 9 Tuesday, 29th October MKZ 1 10:45 - 12:30 Discussion 3: Searle 10 Tuesday, 5th November MKZ 1 10:45 - 12:30 Lecture 7: Quantifying Uncertainty Thursday, 7th October Practical 5: Probabilistic Reasoning (1) 11 Wednesday, 13th November MKZ 1 08:45 - 10:30 Discussion 4: Ethics of AI 12 Wednesday, 20th November Aula 08:45 - 10:30 Lecture 8: Probabilistic Inference Thursday, 21st November Practical 6: Probabilistic Reasoning (2) 13 Lecture 9: Summary and Conclusion TBA Final Exam 36 39 Is there a course text? I Russell, S. & Norvig, P. (2020). Artificial Intelligence: A Modern Approach (4th ed.). Prentice Hall, NJ. I Russell, S. & Norvig, P. (2009). Artificial Intelligence: A Modern Approach (3rd ed.). Prentice Hall, NJ. 37 39 Your task for next week Read: Russell & Norvig chapter 1, pages 1-29. This covers some of the issues we touched on today, and some we will cover next week. 38 39