November 30, 2011 - Aritificial Intelligence.docx

5 Pages
89 Views
Unlock Document

Department
Media, Information and Technoculture
Course
Media, Information and Technoculture 2500A/B
Professor
John Reed
Semester
Fall

Description
Artificial Intelligence November 30, 2011 I. Narrow AI assist us - Early AI/intelligence tests - Top-down AI - Bottom-up AI II. General AI exceeds us III. Kurzweil - Genetic engineering - Nanotech + Nano robotics IV. Evolution beat down V. Kaczynski 1961 Bell Labs – Computer could sing daisy Blurring natural and artificial intelligence I. Narrow AI - Narrow AI (assist us) and General AI (exceeds us) - Narrow does not match or exceed humans, it is specialized, it is extremely intelligent in a very narrow way Early AI - Babbage’s analytical engine: 1850’s - First programmable device, about the size of the room and weighed tons - Never built but it was theorized - Would have been powered by steam - Input: Cards, processing unit - Output: Printer Eniac 1946 (first electronic computer) - You have to program every possible move - Doesn’t think in real time Field was established in 1959 - Might be very intelligent but in a very narrow way B. Intelligence Tests (starting in the 30’s and 40’s, how do we know when something has become intelligent?) - Turing Test (1950, first intelligence test, cybernetics Norbert Weiner) - Drinking game, changed it to apply to the computer, ask questions to a computer and ask questions to a human, when you can’t tell the difference between who is answering, than it has become intelligent Eliza (1966, second intelligence test) - Responses on based keywords - Programmed to look for key words and respond based - Draws user into certain patterns, it is easily fooled - All these tests prove intelligence but not consciousness C. Top down AI - One approach to narrow AI - Program all possible information in it - Checkers (1959) picks within those possibilities, but it requires a lot of research and front loading Shakey (1966-1972) - Put in all the possibilities into shakey and then it reasons from that - Problem: Easily confound, only know what is programmed, it doesn’t have general knowledge, general AI is the goal - Top down programming, programming every possibility in advance Expert Systems - Has a very narrow but deep sense of knowledge - Microsoft’s help file - Duplicate one or more human experts - Spell check, expert systems - Problem: They are very brittle, not very flexible - These expert systems, might be very brilliant but in a very inflexible way Broad and shallow - Use children as an example, acquiring tons of knowledge over a period of time - Programmed computers the way that children act, to train them to play - Build frame, broad and shallow - Example: Facial identification of iPhoto, you give it certain frames and it will slowly evolve and fill in the gaps - Problem: Lacks common sense, enormous amount of top down loading into these systems, abandoned in the early 90’s CYC PROJECT (teach the system common sense) - This is Abraham Lincoln, teach it common sense, when he was in Washington, his left foot was in Washington D. Bottom up AI - The idea that simple patterns can create complex systems - Looks at the body on a cellular and chemical level as a self-replicating system - Goes back to the cybernetic system - Brain doesn’t function as a top down computer, the brain is more like an anthill, a constantly evolving, changing system, order within the chaos (THE BRAIN IS GRASS) - Evolutionary model, self-adjusting, changing - Not pre planned, not functioning in a standardized system - It is emergent, very much about order emerging from chaos, going with the flow and seeing where things go from there - Stabilization is very important - Simple patterns will create complex systems if you give them time to evolve Cellular automata (1940s) - Simple patterns can create complex systems - The game of life, the complex evolutionary system that self creates and self organizes, no way to predict what its going to do, totally against fordist ways, it was a post industrial aspect - On/Off states and its state of the neighbors - A system that can self organize, order within the chaos, does not follow along the Taylorist, industrial model of predictability Virtual Evolution (1994): - Evolution of very simple agents in an artificial environment - Operated through a genetic algorithm - Literally a computer algorithm Swarm Intelligence - Collective order out of chaos, you see this at work when you drop sand - Somewhat predictable, even though it is made up of billions of particles, it will usually form a hill - You have simple agents (grain of sand, bird, fish) that continually come toget
More Less

Related notes for Media, Information and Technoculture 2500A/B

Log In


OR

Join OneClass

Access over 10 million pages of study
documents for 1.3 million courses.

Sign up

Join to view


OR

By registering, I agree to the Terms and Privacy Policies
Already have an account?
Just a few more details

So we can recommend you notes for your school.

Reset Password

Please enter below the email address you registered with and we will send you a link to reset your password.

Add your courses

Get notes from the top students in your class.


Submit