Presentation Outline : Overview
-
Steve Austin & Peter Norvig
-
Turing’s Imitation Game
-
About Agents
-
Engineering AMEE
-
Summary
Steve Austin & Peter Norvig
-
Steve Austin
-
"We can rebuild him. We have the technology. We can make him better than he was. Better…stronger…faster." — The Six Million Dollar Man, 1973
-
The technology of the day
-
The power of a machine
-
The brain of a human
-
-
Peter Norvig
-
"More data beats clever algorithms, but better data beats more data." — Peter Norvig
-
"How to Write a Spell Corrector" — Peter Norvig (2007) : http://norvig.com/spell-correct.html
-
"Artificial Intelligence : A Modern Approach" — Russell & Norvig (1994)
-
Turing’s Imitation Game
-
Alan Turing
-
"Are there imaginable digital computers which would do well in the imitation game?" — Alan Turing, 1950
-
About Agents
-
Intelligent Agent
-
"Anything that can be viewed as perceiving its environment through sensors and acting upon that environment through actuators" — Russell & Norvig, "Artificial Intelligence: A Modern Approach" (1994)
-
-
Autonomous Agents
-
"Software entities that carry out operations on behalf of a user with independence and employ knowledge of the user’s goals." — a multiply cited statement in an IBM white paper no longer accessible
-
-
PAGE
-
Percepts, Actions, Goals, Environment
-
Percepts : Something that is perceived.
-
Actions : Something that is done; an act.
-
Goals : A desired result.
-
Environment : Conditions under which you operate.
-
-
Classes of Agents
-
Simple reflex agent : door robot
-
Model-based agent : vacuum robot
-
Goal-based agent : find this vacuum robot for me online
-
Utility agent : buy this vacuum robot at the best price
-
Learning agent : remember where you get the best price on vacuum robots
-
-
Intelligent Autonomous Agents
-
Imitating users
-
Carrying out tasks
-
Using percepts, actions, goals, and environment
-
Via reflex, model, goal, utility, learning
-
Engineering AMEE
AMEE : The Autonomous Maze Environment Explorer
-
AMEE : What is it?
-
AMEE can successfully navigate any 2-dimensional maze of arbitrary size.
-
Agent Class: Goal-based
-
Percepts: Various kinds of doors
-
Actions: Affordances to view, turn, and move
-
Goals: Escape!
-
Environment: XML messages via HTTP protocol
-
-
AMEE : How do we build it? : https://github.com/mamund/2021-02-dagstuhl/blob/master/alps/asd/maze-alps.svg
-
AMEE’s algorithm (Model) : https://github.com/mamund/2021-02-dagstuhl/blob/master/images/2021-02-13-maze-bot-model.png
-
Coding AMEE (Sensors & Actuators) : https://github.com/mamund/rwa/blob/master/Maze/the-bot/the-bot.js
Summary
-
The Future of AMEE
-
AMEE’s Environment : Include rewards (added points) and dangers (lost points)
-
Agent Class : Level up to utility[L4] (scoring) and learning[L5] (improving score)
-
-
Imitating AMEE in the real world
-
Navigation
-
Selection
-
Acquiring
-
Remembering
-
Learning
-
One more thing…
-
Rene Descartes
-
"It is impossible for a machine to have enough different organs to make it act in all the contingencies of life in the way in which our reason makes us act." — Rene Descartes, Discourse on Method (1637)
-
-
We need lots of AMEEs!
Thank you!
-
"We can built it, we have the technology…"
-
"Imaginable digital computers…"
-
PAGE & Agent Classes
-
Engineering AMEE
-
"Enough different AMEEs …"