Upload
hector-marsh
View
218
Download
1
Tags:
Embed Size (px)
Citation preview
Artificial Intelligence:Artificial Intelligence:
Human vs. MachineHuman vs. Machine
Professor Marie desJardins
CMSC 100
November 5, 2009
2
Memory is at the Core (Literally)
Remember Hal? “Open the pod bay door, Hal.” “My mind is going...”
Memory is at the core of our being (and a computer’s)
...but our memories look very different!
The first magnetic core memory [www.columbia.edu/acis/history] thebrain.mcgill.ca
3
Overview
What is AI? (and why is it so cool?) AI: Past and Present
History of AI AI Today
Computational vs. Biological Memory The Skeptics Speak
What is AI?What is AI?
5
AI: A Vision
Could an intelligent agent living on your home computer manage your email, coordinate your work and social activities, help plan your vacations…… even watch your house while you take those well planned vacations?
6
Represent and store knowledge Retrieve and reason about knowledge Behave intelligently in complex environments Develop interesting and useful applications Interact with people, agents, and the environment
Main Goals of AI
7
ComputerScience &
Engineering
AI
Mathematics
CognitiveScience
Philosophy
Psychology Linguistics
BiologyEconomics
Foundations of AI
8
Big Questions
Can machines think? If so, how? If not, why not? What does this say about human
beings? What does this say about the mind? And if we can make machines think,
should we?
History of AIHistory of AI
11
History
12
ELIZA: A program that simulated a psychotherapist interacting with a patient and successfully passed the Turing Test
Implemented at MIT during 1964-1966 by Joel Weizenbaum
First script was DOCTOR Simple pattern-matching Each pattern had an associated reply which might include
bits of the input (after simple transformations (my your) Weizenbaum was shocked at reactions
Psychiatrists thought it had potential People unequivocally anthropomorphized
Let me tell you my problems... Many thought it solved the natural language problem!!
Early AI: Eliza
13
Eliza Speaks
I am the psychotherapist. Please, describe your problems. i'm having a hard time with my students Why do you say that? well, they keep asking me for money Maybe your plans have something to do with this. what plans? Is it because of your life that you say what plans? No, it's because I didn't know what you were talking about Is it because you didn't know what I was talking about that you
came to me? no, it's because i wanted to see if you could pass the turing test Does it bother you that you wanted to see if I could really pass the
turing test? ….. How do you reconcile problems at school? This is a waste of time. Good bye My secretary will send you a bill.
14
Early AI: Parry
Kenneth Colby modeled a paranoid using the same techniques as Eliza, around 1968
PARRY has basic emotions. If it gets angry, its replies become more hostile
In a Turing test, expert psychiatrists were unable to distinguish PARRY’s ramblings from those of real paranoids
15
PARRY Speaks
Good afternoon. My name is Doctor Jones. Pleased to meet you. How are you? All right I guess. Could you tell me your name? Pat Smith. I though you might have some problems. I've been upset lately. By what? People get on my nerves sometimes. Tell me more. Do you know anything about bookies? ...
16
P: I don't understand your motives.
E: Why don't you understand my motives?
P: Shouldn't I be?
E: Why do you say that just now?
P: That's just the way it is.
E: Suppose that's the way it were not.
P: So?
E: I am not sure I understand you fully.
P: You harp on the same questions too much.
E: What does that similarity suggest to you?
P: I don't get you.
E: Do you wish that you did get I?
17
Turing Test
Three rooms contain a person, a computer, and an interrogator
The interrogator can communicate with the other two by “teleprinter” (or, say, AIM)
The interrogator tries to determine which is the person and which is the machine
The machine tries to fool the interrogator into believing that it is the person
If the machine succeeds, then we conclude that the machine can think
18
The Loebner Contest
A modern version of the Turing Test, held annually, with a $100,000 cash prize
Hugh Loebner was once director of UMBC’s Academic Computing Services (née UCS, lately OIT)
http://www.loebner.net/Prizef/loebner-prize.html Participants include a set of humans, a set of computers,
and a set of judges Scoring
Rank from least human to most human Highest median rank wins $2000 If better than a human, win $100,000 (Nobody yet…)
2008 winner: Elbot
19
What’s Easy and What’s Hard?
It’s been easier to mechanize many of the high-level tasks we usually associate with “intelligence” in people e.g., symbolic integration, proving theorems, playing
chess, medical diagnosis It’s been very hard to mechanize tasks that lots of animals
can do walking around without running into things catching prey and avoiding predators interpreting complex sensory information (e.g., visual, aural, …) modeling the internal states of other animals from their behavior working as a team (e.g., with pack animals)
Is there a fundamental difference between the two categories?
AI TodayAI Today
21
Who Does AI?
Academic researchers (perhaps the most Ph.D.-generating area of computer science in recent years) Some of the top AI schools: CMU, Stanford, Berkeley, MIT, UIUC,
UMd, U Alberta, UT Austin, ... (and, of course, UMBC!)
Government and private research labs NASA, NRL, NIST, IBM, AT&T, SRI, ISI, MERL, ...
Lots of companies!
22
A sample from the 2008 International Conference on Innovative Applications of AI: Event management (for Olympic equestrian competition) Language and culture instruction Public school choice (for parents) Turbulence prediction (for air traffic safety) Heart wall abnormality diagnosis Epilepsy treatment planning Personalization of telecommunications services Earth observation flight planning (for science data) Crop selection (for optimal soil planning)
Applications
23
Here are some example applications: Computer vision: face recognition from a large set Robotics: autonomous (mostly) automobile Natural language processing: simple machine translation Expert systems: medical diagnosis in a narrow domain Spoken language systems: ~2000 word continuous speech Planning and scheduling: Hubble Telescope experiments Learning: text categorization into ~1000 topics User modeling: Bayesian reasoning in Windows help (the infamous
paper clip…) Games: Grand Master level in chess (world champion), checkers,
backgammon, etc. Breaking news (8/7/08) - MoGo beats professional Go player
What Can AI Systems Do Now?
24
Robotics
SRI: Shakey / planning sri-shakey.ram SRI: Flakey / planning & control sri-Flakey UMass: Thing / learning & control
umass_thing_irreg.mpegumass_thing_quest.mpegumass-can-roll.mpeg
MIT: Cog / reactive behaviormit-cog-saw-30.movmit-cog-drum-close-15.mov
MIT: Kismet / affect & interactionmit-kismet.movmit-kismet-expressions-dl.mov
CMU: RoboCup Soccer / teamwork & coordinationcmu_vs_gatech.mpeg
25
DARPA Grand Challenge
Completely autonomous vehicles (no human guidance) Several hundred miles over varied terrain First challenge (2004) – 142 miles
“winner” traveled seven(!) miles
Second challenge (2005) – 131 miles Winning team (Stanford) completed
the course in under 7 hours Three other teams completed the
course in just over 7 hours
Onwards and upwards (2007) Urban Challenge Traffic laws, merging, traffic
circles, busy intersections... Six finishers (best time: 2.8 miles in 4+ hours)
26
Art: NEvAr
Use genetic algorithms to evolve aesthetically interesting pictures
See http://eden.dei.uc.pt/~machado/NEvAr
27
ALife: Evolutionary Optimization
MERL: evolving ‘bots
28
Human-Computer Interaction: Sketching
Step 1: Typing Step 2: Constrained handwriting Step 3: Handwriting recognition Step 4: Sketch recognition (doodling)! MIT sketch tablet
29
Driving: Adaptive Cruise Control
Adaptive cruise control and pre-crash safety system (ACC/PCS)
Offered by dozens of makers, mostly as an option (~$1500) on high-end models
Determines appropriate speed for traffic conditions
Senses impending collisions and reacts (brakes, seatbelts)
Latest AI technology: automatic parallel parking!
30
AxonX
Smoke and fire monitoring system
31
Rocket Review
Automated SAT essay grading system
32
What Can’t AI Systems Do (Yet)?
Understand natural language robustly (e.g., read and understand articles in a newspaper)
Surf the web (or a wave) Interpret an arbitrary visual scene Learn a natural language Play Go well √ Construct plans in dynamic real-time domains Refocus attention in complex environments Perform life-long learning
Computational vs. Biological Memory
34
How Does It Work? (Humans)
Basic idea: Chemical traces in the neurons of the brain
Types of memory: Primary (short-term) Secondary (long-term)
Factors in memory quality: Distractions Emotional cues Repetition
35
How Does It Work? (Computers)
Basic idea: Store information as “bits” using physical processes (stable
electronic states, capacitors, magnetic polarity, ...) One bit = “yes or no”
Types of computer storage: Primary storage (RAM or just “memory”) Secondary storage (hard disks) Tertiary storage (optical jukeboxes) Off-line storage (flash drives)
Factors in memory quality: Power source (for RAM) Avoiding extreme temperatures
Size Speed
36
Measuring Memory
Remember that one yes/no “bit” is the basic unit Eight (23) bits = one byte 1,024 (210) bytes = one kilobyte (1K)*
1,024K (220 bytes) = one megabyte (1M) 1,024K (230 bytes) = one gigabyte (1G) 1,024 (240 bytes) = one terabyte (1T) 1,024 (250 bytes) = one petabyte (1P) ... 280 bytes = one yottabyte (1Y?)
* Note that external storage is usually measured in decimal rather than binary (1000 bytes = 1K, and so on)
39
Moore’s Law
Computer memory (and processing speed, resolution, and just about everything else) increases exponentially
40
Showdown
Computer capacity: Primary storage: 64GB Secondary storage: 750GB
(~1012) Tertiary storage: 1PB? (1015)
Computer retrieval speed: Primary: 10-7 sec. Secondary: 10-5 sec.
Computing capacity: 1 petaflop (1015 floating-point instructions per second), very special purpose
Digital Extremely reliable Not (usually) parallel
Human capacity: Primary storage: 7 ± 2 “chunks” Secondary storage: 108432 bits??
(or maybe 109 bits?)
Human retrieval speed: Primary: 10-2 sec Secondary: 10-2 sec
Computing capacity: possibly 100 petaflops, very general purpose
Analog Moderately reliable Highly parallel
More at movementarian.com
41
It’s Not Just What You “Know”
Storage Indexing Retrieval Inference Semantics Synthesis
...So far, computers are good at storage, OK at indexing and retrieval, and humans win on pretty much all of the other dimensions
...but we’re just getting started Electronic computers were only invented 60 years ago! Homo sapiens has had a few hundred thousand years to evolve...
The Skeptics SpeakThe Skeptics Speak
43
Mind and Consciousness
Many philosophers have wrestled with the question: Is Artificial Intelligence possible?
John Searle: most famous AI skeptic Chinese Room argument
Is this really intelligence?
?
!
44
What Searle Argues
People have beliefs; computers and machines don’t. People have “intentionality”; computers and machines don’t. Brains have “causal properties”; computers and machines
don’t. Brains have a particular biological and chemical structure;
computers and machines don’t.
(Philosophers can make claims like “People have intentionality” without ever really saying what “intentionality” is, except (in effect) “the stuff that people have and computers don’t.”)
45
Let’s Introspect For a Moment...
Have you ever learned something by rote that you didn’t really understand?
Were you able to get a good grade on an essay where you didn’t really know what you were talking about?
Have you ever convinced somebody you know a lot about something you really don’t?
Are you a Chinese room??
What does “understanding” really mean? What is intentionality? Are human beings the only entities
that can ever have it? What is consciousness? Why do we have it and other
animals and inanimate objects don’t? (Or do they?)
46
Just You Wait...