W1 - introduction to intelligent systems st.pdf

ECE 579/CIS 579 Intelligent systems Yi Lu Murphey ECE 579: Intelligent Systems -- Introduction Instructor: Professor Yi Lu Murphey Dept. Of Electrical & Computer Engineering The University of Michigan-Dearborn...

ECE 579/CIS 579 Intelligent systems Yi Lu Murphey ECE 579: Intelligent Systems -- Introduction Instructor: Professor Yi Lu Murphey Dept. Of Electrical & Computer Engineering The University of Michigan-Dearborn [email protected] 1 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey What is intelligence  Is it just a name for a collection of distinct and unrelated abilities?  Can intelligence be inferred from observable behavior, or does it require evidence of a particular internal mechanism?  How is knowledge represented in the nerve tissue of a living being  Should we pattern an intelligent computer program after what is known about human intelligence? Example: computer vision vs. human vision, ANN vs. NNN.  Is it possible to achieve intelligence on a computer?  Is it possible for a computer to be smarter than human?(Chess play: deep blue vs. Kosporove, IBM Watson wining Jeopardy)  Does an intelligent entity require the richness of sensation and experience that might be found only in a biological existence?  What human intelligence have been successfully achieved by a computer? what have not?  More discussions can be found in many AI-related text books and research papers.  All these questions are open for debate! 2 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey Definition of Intelligence  Definition from Merriam-Webster Dictionary: – the ability to learn or understand or deal with new or trying situations – the ability to apply knowledge to manipulate one's environment or to think abstractly as measured by objective criteria (as tests)  In this class, we give the following definition: – Intelligence is the ability to learn or understand new knowledge and/or skills and apply knowledge/skill to solve problems. 3 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey Study of intelligence  The study of intelligence is one of the oldest disciplines  For over 2000 years, philosophers have tried to understand how seeing, learning, remembering, and reasoning could, or should, be done.  Artificial Intelligence, George F. Luger. Addison-Wesley, 4th Edition, 2001 – This text book gave a very detailed discussion on this topic. 4 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey Artificial intelligence: Based on Artificial Intelligence– A modern Approach By Stuart Russell, Peter Norvig, second edition  A discipline that has many different definitions (from 8 textbooks) – The branch of computer science that is concerned with the automation of intelligent behavior[Luger, 2001] – The study of the computations that make it possible to perceive, reason, and act.[Winston, 1992] – The study of mental faculties through the use of computational models.[charniak and McDermott, 1985] – The exciting new effort to make computers think … machines with minds, in the full and literal sense. [Haugeland, 1985] – The automation of activities that we associate with human thinking, activities such as decision-making, problem solving, learning, …”[Bellman, 1978] – The art of creating machines that perform functions that require intelligence when performed by people[Kurzweil, 1990] – The study of how to make computers do things at which, at the moment, people are better[Rich and Knight 1991] 5 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey What is AI?  To summarize all these definitions, we define AI as a study about how to make a computer to have human’s – Learning and thought processes, – reasoning processes, and – rational behaviors 6 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey What is an intelligent system?  An intelligent system is a computer program that can learn to perform specific tasks that require intelligence when performed by people. – Can any engineering approach to a problem be considered as an intelligent system? – An intelligent system must have the capability to learn to perform specific tasks and generalize to new cases.  An intelligent system has the capability of perceive, learn, reason, or/and act.  An intelligent system that automates certain intelligent behaviors.  An intelligent system is a computer program that should – think and act like humans or – think and act rationally 7 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey Categories of Intelligent Systems  The study of intelligent systems has been undertaken in two broad categories: – The first and more popular category falls under the label of artificial intelligence. – The second category is termed inductive inference or machine learning 8 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey Intelligent systems: State of Art  Automatic inspection of electronic boards  Fuzzy controllers  Advanced signal diagnostics  Digital image indexing and retrieval  Robotics  Engineering diagnostics  Medical diagnostics  Stock market prediction systems  Intelligent information retrieval systems: vedio, text, speech, …  Intelligent manufacturing systems: – Robots in assembly plants – Classification of defect products – Supply chain management  Intelligent vehicle power management  Risk analysis and management – Forecasting, customer retention, improved underwriting, quality control, competitive analysis  Fraud detection and management  Text mining (news group, email, documents)  Web analysis – Intelligent query answering  Chatbot: text based or spoken conversation – A Chatbot is an artificial person, animal or other creature which holds conversations with humans. 9 – Eg. ChatGPT, ECE 579/CIS 579 Intelligent systems Yi Lu Murphey Machine Learning Approaches to Vehicle Power Management An innovative machine learning framework for the generation of intelligent vehicle power controllers Fuel reduction NNdc: predicting current driving cycles NNdt: predicting current driving trends NNom: predicting current operating modes NNds: predicting driving styles A fuzzy learning algorithm that automatically generates a fuzzy knowledge base containing optimized fuzzy rules and fuzzy membership functions A fully developed and optimized intelligent power controller for a conventional vehicle A fully developed and optimized intelligent power controller for a parallel hybrid vehicle system 10 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey Text Document Mining using Machine Learning Query document: an example of customer comments ENG RUNS ROUGH & GETTING POOR FUEL MILEAGE Tx2DC System (Text-to-Diagnostic Code) Customer Concern (CC) code: Code Mapping CSC/OASIS code: 11 Automotive diagnostic text characteristics:  Examples of auto problem description – AM TO START ENGINE – HESITATES AT TIMES WHEN ACCEL. FROM A STOP *DIESEL* – SPW CLAIM BATTERY DEAD SHORT – VEH BUCKS AT HIGHWAY SPEEDS – ENG RUNS ROUGH & GETTING POOR FUEL MILEAGE – C S CK TRANS FOR LEAK OIL SEAL LEAKING – C S CHECK FOR GAS CAP BROKEN – C S HAS BAD FLUID LEAKS ENGINE OR TRANS CK AND ADVISE; – CUST STATES THE RIGHT HEADLIGHT APPEARS TO BE DIMMER THAN THE OTHER SIDE – CHANGE OIL, FILTER AND LUBRICATION PREFERRED CUSTOMER 5TH OIL CHANGE, NO CHARGE – CUSTOMER STATES TIRE FLAT PERFORM 22500 MILE SERVICE – CUSTOMER STATES R/F TURNSIGNAL LENS IS BROKEN 12 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey Challenges in Mining Diagnostic Text  Data are often ill-structured – HESITATES AT TIMES WHEN ACCEL. FROM A STOP *DIESEL*  Text descriptions often do not follow the English grammar  many irrelevant terms occur frequently – C S HAS BAD FLUID LEAKS ENGINE OR TRANS CK AND ADVISE;  many self-invented acronyms and shorthand are used in descriptions – C S, CUST, SPW  Large variations of typos: – Rugh, ruf, engen, ingen, enginne, accellate, acelerate, accl, etc 13 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey How does ChatGPT interpret these descriptions?  HESITATES AT TIMES WHEN ACCEL. FROM A STOP *DIESEL*  chatGPT; – The given statement appears to describe a behavior related to acceleration in a diesel engine. Specifically, it suggests that there are instances when the engine hesitates or experiences a delay when accelerating from a complete stop. The mention of "diesel" indicates that this is a characteristic or issue specific to diesel engines. This hesitation might manifest as a momentary delay or lack of immediate response during acceleration. If you encounter this issue with a diesel vehicle, it may be advisable to have the engine checked by a mechanic to identify and address the underlying cause. 14 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey How does ChatGPT interpret these descriptions?  C S HAS BAD FLUID LEAKS ENGINE OR TRANS CK AND ADVISE;  ChatGPT: – The sentence "C S HAS BAD FLUID LEAKS ENGINE OR TRANS CK AND ADVISE" is a brief note often found in the context of automotive or mechanical repair. Let's break it down:  "C S": This likely stands for "customer says" or "client states," indicating that the information following this abbreviation is based on what the customer has reported or complained about.  "HAS BAD FLUID LEAKS": The customer has reported significant leaks of fluids. This could be from either the engine or the transmission.  "ENGINE OR TRANS": The leaks are associated with either the engine or the transmission.  "CK AND ADVISE": The mechanic is instructed to check (inspect) the vehicle and provide advice or recommendations on how to address the identified fluid leaks.  In summary, the sentence is a concise instruction for a mechanic to investigate and assess a vehicle due to reported significant fluid leaks, specifying that the leaks could be from either the engine or the transmission. The mechanic is then expected to check the vehicle and provide guidance on how 15 to address the issue. ECE 579/CIS 579 Intelligent systems Yi Lu Murphey Real-time Vehicle Detecting and Tracking  Fast road detection  Robust vehicle detection  Scene monitoring  Vehicle tracking 16 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey Passenger Seat Occupant Classification  In-vehicle seat occupant classification is critical in airbag safety control. – many traffic fatalities involving inappropriate airbag deployments.  We have developed a suite of low-cost and fast speed computer vision algorithms that can distinguish 5 different classes of occupants in the passenger seat: adult Child in a child seat Child Empty seat Empty child seat 17 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey People Detection and Intent Classification  Motion detection  Human-centered segmentation  Human detection  Intent classification – Walking vs. running – Threat vs. non-threat 18 ECE 579/CIS 579 Intelligent systems Yi Lu Murphey An intelligent system must have the following components  Mechanisms to learn domain knowledge: – machine learning algorithms  knowledge representation – the study of how to put knowledge into a form that a computer can reason with – express knowledge in computer-tractable form, such that it can be used to help an intelligent system perform tasks well.  Abilities to perform complicated tasks based on reasoning and inference within reasonable time. 19 ECE 579 Intelligent systems Yi Lu Murphey Garry Kasparov’s chess match against Deep Blue (1)  Deep Blue: IBM supercomputer Deep Blue  Garry Kasparov: World Chess Champion.  The first match – February 1996 in Philadelphia, Pennsylvania. – Kasparov won the match 4–2, losing one game, drawing in two and winning three.  The second match – May, 1997 in New York City – has been called "the most spectacular chess event in history“ – Deep Blue won 3½–2½.  Kasparov proposed another rematch for ’98, but IBM would have none of it. – The company dismantled Deep Blue, which never played 20 chess again. ECE 579 Intelligent systems Yi Lu Murphey Garry Kasparov’s chess match against Deep Blue (2)  Technology behind IBM Deep Blue – an IBM RS/6000 SP parallel supercomputer equipped with chess specific coprocessors.  The Deep Blue system is capable of – examining 200 million moves per second or – 50 billion positions in the three minutes nominally allotted for a single move in a chess game.  Deep Blue had used its incredible processing power and the accumulated knowledge of computer scientists and chess champions to engage in something resembling "thought."  Major techniques used in Deep Blue – advanced search algorithms, large-scale parallelization, application specific hardware  The IBM researchers took what they learned from building Deep Blue -- and from lessons learned during the match -- to tackle other complex problems. – Computational Finance and Data Mining 21 ECE 579 Intelligent systems Yi Lu Murphey IBM Watson won Jeopardy! (1)  Watson (2011) – is an artificial intelligence computer system capable of answering questions posed in natural language – was developed in IBM's DeepQA project by a research team led by principal investigator David Ferrucci. – was named after IBM's first president, Thomas J. Watson. – Has an avatar inspired by the “IBM’s smarter planet” symbol 22 ECE 579 Intelligent systems Yi Lu Murphey IBM Watson won Jeopardy! (2)  In 2011, Watson competed on the quiz show Jeopardy!, – the show's only human-versus-machine match-up to that date. – In a two-game, combined-point match, broadcast in three Jeopardy! episodes February 14–16, Watson beat » Brad Rutter, the biggest all-time money winner on Jeopardy!, and » Ken Jennings, the record holder for the longest championship streak (74 wins). – Watson received the first prize of $1 million, while Ken Jennings and Brad Rutter received $300,000 and $200,000, respectively. » Jennings and Rutter pledged to donate half their winnings to charity » IBM divided Watson's winnings between two charities. 23 ECE 579 Intelligent systems Yi Lu Murphey IBM Watson won Jeopardy! (3)  Watson consistently outperformed its human opponents on the game's signaling device  Watson had trouble responding to a few categories, notably those having short clues containing only a few words.  Watson was not connected to the Internet during the game. 24 ECE 579 Intelligent systems Yi Lu Murphey Watson Technology  Watson, IBM's natural-language processing supercomputer, – Its appearance is surprisingly cute » "He" appears as a black rectangle with a globe avatar sitting between his rivals, Jeopardy champs Ken Jennings and Brad Rutter.  Watson "understands" language as humans naturally speak it. – That's no easy feat for a computer, since Human language is full of subtleties, irony and words with multiple meanings. – IBM has been working on the project for several years.  Watson was loaded to an IBM machine with 200 million pages of structured and unstructured content consuming four terabytes of disk storage, including the full text of Wikipedia.  Watson used that data to analyze contextual clues and figure out how words relate to each other. 25 ECE 579 Intelligent systems Yi Lu Murphey Performance analysis  Watson processes the questions by considering many factors, ranging from straightforward keyword matching to more complex challenges – Homonyms » the bark of a tree is not the same as a dog's bark) – statistical paraphrasing » "Big Blue" is the same thing as "IBM“ – The final margin seemed to come down to who could buzz in the fastest. 26 ECE 579 Intelligent systems Yi Lu Murphey Why Watson was considered a landmark in computational intelligence?  In design of Watson – IBM researchers say they avoided merely loading Watson up with countless databases because they want the technology to work for any person or company looking for quick answers in any application field  Possible applications of Watson technology – health care area. » Watson could help doctors more accurately diagnose patients – business applications include » online self-service help desks, tourist information centers, customer hotlines and more, etc. 27 ECE 579 Intelligent systems Yi Lu Murphey What other IBM Watson systems  IBM Watson is a technology platform that uses natural language processing and machine learning to reveal insights from large amounts of unstructured data – Examples: » Answer customers' most pressing questions » Quickly extract key information from all documents » Discover insights, patterns and relationships across data  IBM has developed – Watson Ecosystems for building business applications using Watson technologies – Watson Health for health care – Watson API for software developers – Watson Hackathon – Chef Watson  https://www.ibm.com/watson/ – IBM Watson is AI for smarter business – Last visited in August 30, 2023 28  ECE 579 Intelligent systems Yi Lu Murphey Chef Watson - cognitive cooking (1)  Chef Watson is part of IBM’s mission to develop cognitive computing applications – that can help people discover new ideas, from creating surprising new recipes to improving medical research.  It has helped thousands of home chefs create delicious meals.  You can use it as a cooking assistant 29 ECE 579 Intelligent systems Yi Lu Murphey IBM Watson Health: current status (1)  Since 2012, Watson for Oncology has undergone intensive medical training by oncologists at Memorial Sloan Kettering Cancer Center (MSK) in New York to learn key data associated with a patient’s cancer, – such as results from blood tests; pathology and imaging reports detailing the type, size, and location of the tumor; and the presence of genetic mutations.  The program then combs through huge amounts of medical literature to provide evidence-based treatment recommendations for that specific patient.  To date, Watson has been trained to support treatment decisions in lung, breast, colorectal, bladder, cervical, endometrial, esophageal, stomach, ovarian, prostate, and thyroid cancers 30 ECE 579 Intelligent systems Yi Lu Murphey IBM Watson Health: current status (2)  In addition to Watson for Oncology, IBM has launched several other oncology-related offerings aimed at accelerating personalized patient care. – Watson for Genomics, – Watson Genomics from Quest Diagnostics (a Watson-powered genomic-sequencing service), and – IBM Watson for Clinical Trial Matching, which uses natural language processing to increase the efficiency and accuracy of the clinical trial matching process.  This project that was supposed to turn IBM’s computing prowess into a scalable program that could deliver state-of- the-art personalized cancer treatment protocols to millions of patients around the world. 31 ECE 579 Intelligent systems Yi Lu Murphey IBM Watson Health: current status (3)  Although Watson has become more accurate over the years, in cancer diagnostics, erroneous recommendations have more serious consequences than a missed question on Jeopardy!  AI systems like Watson can still have broad applications in medicine. – AI or AI-powered robots excel in performing repetitive tasks with defined steps, such as simple routine surgeries of the eye or hair – analysis of X-rays or other scans, – checking on patients between office visits and handling administrative billing or claims – analyzing clear, structured data such as genetic information. » During a study, Watson was able to identify previously unidentified mutations that proved important to therapeutic recommendations.  Current trend: Human-Computer Symbiosis – AI-Assisted Decision Making in medicine, » doctors may rely on AI systems to analyze patient data and suggest diagnoses or treatment plans, while the doctor’s expertise and judgment guide the final decision. 32 ECE 579 Intelligent systems Yi Lu Murphey AlphaGo: a Google AI system  http://www.inquisitr.com/2873665/alphago-vs-lee-sedol-3-0-google- deepminds-ai-impresses-tech-world/  Google DeepMind Challenge – March 9 ~ 15, 2016.  AlphaGo has made a historical win by beating multiple world champion Lee Sedol – winner of 18 world titles, – famed for his creativity and widely considered to be the greatest player of the past decade.  AlphaGo emerged a surprise 4-1 victor of Lee  The system was invented by DeepMind, co-founded by scientist Demis Hassabis  AlphaGo was programmed to understand and mimic the strategic analysis of expert Go players in the world  AlphaGo has a capability to expand its knowledge. 33 ECE 579 Intelligent systems Yi Lu Murphey Conclusion:  It is clear that computers with human-level intelligence (or better) have already made a huge impact on our everyday lives and would have more on the future course of civilization.  AI could be one of humanity’s most useful inventions 34 ECE 579 Intelligent systems Yi Lu Murphey Study questions  What are the major components in an intelligent system?  Identify these components in these AI systems discussed in this lecture. 35 ECE 579 Intelligent systems Yi Lu Murphey Answers:  Mechanisms to learn domain knowledge: – machine learning algorithms  knowledge representation – the study of how to put knowledge into a form that a computer can reason with – express knowledge in computer-tractable form, such that it can be used to help an intelligent system perform well.  Abilities to perform complicated tasks based on reasoning and inference.  Computational speed 36 ECE 579 Intelligent systems Yi Lu Murphey ChatGPT  ChatGPT (Generative Pre-trained Transformer) is a chatbot launched by OpenAI in November 2022.  It is built on top of OpenAI's GPT-3.5 family of large language models, – It is fine-tuned with both supervised and reinforcement learning techniques.  ChatGPT was launched as a prototype on November 30, 2022, and quickly garnered attention for its detailed responses and articulate answers across many domains of knowledge.  ChatGPT interacts with user in a conversational way.  The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.  Its current version GPT-4 or GPT-4 turbo (8/31/2024) 37 https://en.wikipedia.org/wiki/ChatGPT ECE 579 Intelligent systems Yi Lu Murphey ChatGPT (2)  ChatGPT's versatility and improvisation skills include its ability to – write and debug computer programs; – compose music, teleplays, fairy tales, and student essays; – answer test questions (sometimes, depending on the test, at a level above the average human test-taker); – write poetry and song lyrics; – emulate a Linux system; – simulate an entire chat room; – play games like tic-tac-toe; – simulate an ATM. – ….  ChatGPT is a technology should be used as a human-computer symbiotic relationship – both entities complement each other’s abilities, leading to enhanced 38 problem-solving and decision-making. ECE 579 Intelligent systems Yi Lu Murphey Intelligent Systems -- built upon multi-disciplinary technologies  The disciplines from which AI/intelligent systems inherited ideas, view points, techniques. – Computer science and engineering – Philosophy – Mathematics – Psychology – linguistics 39 ECE 579 Intelligent systems Yi Lu Murphey This class studies  Technologies that can be used to build intelligent systems – Use machine learning to build intelligent systems  Important issues related to the design and building of intelligent systems. 40 ECE 579 Intelligent systems Yi Lu Murphey End 41