Sunteți pe pagina 1din 4

Andrew

Shidel Homework 1

Written Problems:

1.A
Computing Machinery and Intelligence Summary:
The paper start by proposing the question Can machines think?. To answer this he
proposes a game in which a human (A) and a machine (B) both try to convince
another human (C) that they are human. Turing then list some objects and
problems with this scenario. Next, he talks about the various contrary views to the
main question. Turing finishes by describing how a learning machine could be
created. In particular, he writes about a learning machine that imitates a child mind.

Rewriting the Rules of Turing's Imitation Game Summary:
First the idea of artificial intelligence is introduced. Some examples are given, and
the question is asked whether or not these thinks display genuine intelligence. He
then introduces Allen Turing and the Turing test. Next, he goes on to explain some
flaws in the Turing test. He then introduces some alternatives to the Turing test
such as Riedls test.

Facebook invents an intelligence test for machines Summary
This article starts by introducing the test that Facebook created for testing AIs. It
then talks about the goals of Facebook and other companies in relation to AI. Next, it
talks about the potential uses of this test. It finishes by talking about some other
approaches for similar to Facebooks

Q: Turing's article was written more than half a century ago, but still provides a
known benchmark for artificial intelligence. How have technological advancements
since that time affected the validity of his proposals?

A: I do not think that technological advancements have significantly affected the
validity of his proposals. This is simply because his proposals did not rely on any
particular technology, but are instead general statements about intelligence.

Q: If you were interacting online, what would convince you that the entity you are
interacting with is human?

A: Assuming that the AI was designed with the purpose of appearing to be human,
and that the AI was sufficiently successful at this goal, then there would be nothing it
could say that could convince me that it was human.

Q: Do you think the new tests Facebook has created will work reliably? How might
you improve them?

A: Yes and No. They would work reliably if the intent of the AI were to solve an
arbitrary word problem. They would not work reliably if the intent was to create a

human-like AI since they do not test for qualities such as personality, emotional
understanding, and real world intuition. To address this, I would create questions
based more off of natural human interactions.

Q: How do you think things will advance in the next ten years?

A: I think that as computers become sufficiently powerful, more general solutions
such as Evolutionary Algorithms and Neural Networks will become more dominant.


1.B.a:

Knowledge Base:
Marbles in each basket.

State:
The number of marbles in each basket.

Rules:

Action
Precondition
move 1->2
len(1)>=len(2)
move 1->3
len(1)>=len(3)
move 2->1
len(2)>=len(1)
move 2->3
len(2)>=len(3)
move 3->1
len(3)>=len(1)
move 3->2
len(3)>=len(2)


State Representation:

State representation can be done as an array if integers. Each basket will get an
index in the array, and the value of the integer will be the number of marbles in the
basket.

Initial State:

For an initial state could be in the form: [5,6,7]. This would mean that basket 1 has
5 marbles, basket 2 has 6, and basket 3 has 7.


Goal/Termination:
All baskets have an equal number of marbles, or no legal moves can be made.

1.B.b


Knowledge Base:
The value of each square.
The Dimensions of the maze.

State:
The current position.

Rules:

Action
Precondition
move down by grid[x][y]
y+grid[x][y] < n
move up by grid[x][y]
y-grid[x][y] >= 0
move right by grid[x][y]
x + grid[x][y] < n
move left by grid[x][y]
x - grid[x][y] >= 0


State Representation:

State representation can be done as a 2D array of integers. The first dimension
could be the column, and the second dimension could be the row. The destination
square could be marked with -1. The position in the grid would be an x,y pair.

Initial State:

The initial state shown in the homework would be

[[3,4,1,3,1],[3,3,3,-1,2],[3,1,2,2,3],[4,2,3,3],[4,1,4,3,2]]


Goal/Termination:
Grid[x][y] == -1










1.C


Agent Type
Nest
Thermostat

Performance
Measure
Temperature
of house stays
close to
desired
settings

Environment

Actuators

Users
interactions.
Weather
conditions In
house.

Display on
Thermometer.
device.
Humidity
Heating
sensor.
and/or cooling
systems.



Task
Observable Agents Deterministic Episodic
Environment
Nest
Thermostat

Partially

Single

Stochastic

Sensors

Static

Discrete

Sequential Dynamic Continuous

S-ar putea să vă placă și