Upload
bp8697818
View
235
Download
0
Embed Size (px)
Citation preview
8/6/2019 Lecture 04 - Fundamental Problems
1/20
Department of Computing Science, Fall 2010
8/6/2019 Lecture 04 - Fundamental Problems
2/20
Artificially Intelligent Behavior
Problems with Classical Artificial Intelligence
By Erik Billing
Department of Computing Science, Fall 2010
8/6/2019 Lecture 04 - Fundamental Problems
3/20
New Homework on Monday!
Write down one topic/question that you find interestingand that you want me to bring up during Lecture 5. The
topic/question should something from Chapter 4 ofPfeifer and Scheier.
Due May 13, 8.00
8/6/2019 Lecture 04 - Fundamental Problems
4/20
Questions
Why is the Traveling Salesman Problem so popular in AI?
Elaborate the difference between agent simulation andclassical simulation
8/6/2019 Lecture 04 - Fundamental Problems
5/20
Traveling salesman problem
A salesman is about to visit a number of cities
He wants to visit each city at least once
He wants to travel the shortest distance possible
A Classical problem in AI
Because its very hard (NP-hard)
The complexity of a exhaustive search if TSP isO(n!)
I.e., the time it takes to search all
combinations increases exponentially with thenumber of cities (n)
8/6/2019 Lecture 04 - Fundamental Problems
6/20
TSP with Ants
Ants run around
randomly searchingfor food
When they find foodthey leave a phero-
mone trace on theirway back to the nest
The trace makes this route more attractive toother ants, which in turn leave their own tracemaking the route even more attractive
Since ants taking a short route will fetch foodfaster than ants taking a longer route, thetrace of the short route will become stronger
8/6/2019 Lecture 04 - Fundamental Problems
7/20
TSP Comparison
TSP by Search Typical example of
Classical AI
It requires arepresentation of theworld
Sequential
Guarantees an optimalsolution
Clear stopping condition
All or nothing
TSP by Ants Typical agent-based
approach
It could be implementedphysically without arepresentation
Parallel
Solution is emergent(may be suboptimal)
No stopping condition
Robust and fault
tolerant
See articleAnt Colony System on the course page
(under extra readings in the literature list)
8/6/2019 Lecture 04 - Fundamental Problems
8/20
Questions
Give another example of the frame problem
8/6/2019 Lecture 04 - Fundamental Problems
9/20
The frame problem
How to keep a model of the world (arepresentation) up to date?
Essentially, anything can happen at any time
How to sport out the relevant possibilitiesform completely irrelevant ones?
R1 The sleeping dog
Everything has to be made explicit R1D1
Consider everything
R2D1
Sort out relevant consequences to consider
8/6/2019 Lecture 04 - Fundamental Problems
10/20
The frame problem - Solution
8/6/2019 Lecture 04 - Fundamental Problems
11/20
Symbol grounding
In GOFAI, intelligence is the manipulation ofsymbols
Symbols
Are defined within a symbol system
Have relations to other symbols
Refer to an object in the outer world
(referent) There are rules of how these symbols may be
combined
How do we connect the symbol with its referent?
8/6/2019 Lecture 04 - Fundamental Problems
12/20
Symbol grounding
Example:
Four symbols {1, 2, +, 3}
Combine the symbols 1 + 2 Produces the symbol 3
A human will interpret the meaning of the
symbol 3 in relation to the specific problem weare trying to solve
8/6/2019 Lecture 04 - Fundamental Problems
13/20
Symbol grounding
Example
Four symbols {CUP, DRINK FULL, EMPTY}
Combine: CUP FULL DRINK Produces: CUP EMPTY
A human can interpret the meaning of this
example and apply it in a specific situation How do we implement this ability in a robot?
8/6/2019 Lecture 04 - Fundamental Problems
14/20
Symbol grounding
Further problems
How do an agent create new symbols?
How do an agent decide what a symbol shouldrefer to before the relation is there?
Remember, symbols are the mechanism ofintelligence, we cant refer to some othermechanism without breaking the PhysicalSymbol System Hypothesis
How do the agent change the ontology
The ontology is the boundary of what theagent can represent (i.e., know)
8/6/2019 Lecture 04 - Fundamental Problems
15/20
Symbol grounding
GIFAI Conclusion
Humans can dosymbol grounding
so it must be
possible
Embodied Conclusion
Humans can not besymbol processing
systems
8/6/2019 Lecture 04 - Fundamental Problems
16/20
Questions
Do true intelligence require a biologicalsystem?
Central processing is a reason for delays
Can we use more powerful processors?
What about real-time processing?
8/6/2019 Lecture 04 - Fundamental Problems
17/20
More Questions
You showed us in class some agents that behavedaccording to some basic rules. When I saw thegraphics that were simulating a population my
question came up:
All the programmed robots/agents used tomodel and predict financial behaviours (stockexchange market, forex exchange market...)
are based in the same concept of "basic" rulesof the ones we saw in class?
8/6/2019 Lecture 04 - Fundamental Problems
18/20
Questions: Intelligence
If I am not wrong, in class, you said that IQ testscould say pretty much if someone is intelligent.
Which aspects of intelligence does an IQ testmeasure? And how does it measures them?
Are a virus intelligent since it can adapt to itsenvironment and maximize its energy?
Why does some defintions of intelligence includethe influence of and adaption to the enviromentand others do not? How can one argue thatintelligence is independent from environmentaladaption?
Why does creativity have to be related toemotions?
8/6/2019 Lecture 04 - Fundamental Problems
19/20
Questions: Searching
Explain Depth-limited search and in whichsituations its useful
What happens to a search problem on a mapwhen you have modern GPS technology, currenttraffics information on roads and so on?
8/6/2019 Lecture 04 - Fundamental Problems
20/20
Consciousness
that the role of consciousness in mental life isvery small, almost frighteningly so. The aspectsof mental life that require consciousness have
turned out to be a relatively minor fraction of thebusiness of the brain''
I think it seems a bit exaggerated. What doesit mean bye "Mental life"?
The book states that consciousness is somethingwe with certainty only can attribute to humans
Where does this statement come from and is
this really a common belief (especially in thescientific society)?