3: Attacks on AI

Question 1

What is The Turing Test?

Answer 1

A test of whether machines can think, especially in terms of the Imitation - allowing for natural language conversations.

Question 2

What are Chatbots?

Answer 2

A computer program or AI that simulates responses in a natural language conversation convincingly similar to a human.

Question 3

What is The Imitation Game?

Answer 3

A test of a machine’s ability to exhibit intelligent behaviour, especially natural language conversations, in such a manner that it becomes indistinguishable from a human.

Question 4

What was the objection to The Imitation Game based on the absence of consciousness in machines?

Answer 4

No machine could feel pleasure, grief, etc.

Turing responded by saying this gave rise to a danger of solipsism, and that the Imitation game exists now - in oral exams

Question 5

What was the objection to The Imitation Game based on the lack of continuity in organic NSes?

Answer 5

the brain does not operate digitally, so thought isn’t possible in digital systems

Turing responded by saying computers can simulate continuous behaviour, e.g. statistically, graphically, numerically - not just binary

Question 6

What was the objection to The Imitation Game based on mathematical objections showing machines cannot perform meta-reasoning?

Answer 6

Godel’s completeness theorem, Halting problem, etc, show that machines cannot do meta-reasoning.

Turing responded by saying we too often give wrong answers ourselves to be justified in being very pleased at the fallibility of machines

Question 7

What were some iterative Imitation Games?

Answer 7

ELIZA, SHRDLU, NICOLAI, Loebner prize

Question 8

What was ELIZA?

Answer 8

An Imitation Game system that used very simple pattern matching, e.g. my and me to your and you.

Question 9

What was SHRDLU?

Answer 9

It had a limited domain, with lessons learnt not being easily abstracted.

Question 10

What was NICOLAI?

Answer 10

Not actually electronic but a person pretending to be one making occasional bad answer. Found out but almost passed Reverse Turing test.

Question 11

What was the Loebner prize test?

Answer 11

An extended Turing test. The issue was that ultimately the Turing Test is easy to pass (and can only be passed) with cheating; cheating to not actually make an AI is better than cheating to make an AI seem like a person.

Question 12

What is the point of the Turing Test?

Answer 12

Ultimately, to examine the question of whether machines can ever think in an operational way: can they carry out the Imitation game?

Question 13

What is Intelligent Classification?

Answer 13

Trying to differentiate between different levels of intelligence, especially humans and computers. An example is ReCAPTCHA.

Question 14

What is neural AI?

Answer 14

???

Question 15

What are neural networks?

Answer 15

AI systems modelled on the structure of the human brain and nervous system.

Question 16

What is symbolic AI?

Answer 16

AI systems using high-level “symbolic” (human-readable) representations of problems, logic, and searches.

Question 17

What is a computer?

Answer 17

An electronic or mechanical device that performs instructions grouped into programs or subroutines, typically using binary data.

Question 18

What does it mean to say that all computers are the same?

Answer 18

???

Question 19

How does a Turing machine work?

Answer 19

Has a tape, and head. Has states and a transition function. Input initially populates tape; head moves to first tape cell. For each cell, repeatedly, the cell is read, and the transition function finds the next state, output to the cell, and movement of the tape head (L/R). Eventually, the TM ends up in an accepting or rejecting state or loops infinitely.

Question 20

Why is a computer a universal machine?

Answer 20

Universal Turing Machine allows you to simulate any other TM. Computers are as you can implement the Universal TM on them.

Question 21

What is weak AI?

Answer 21

Artificial Intelligence which can only process data to give a suitable output but does not possess a mind or thought, and thereby does not actually understand what it is doing. Can be thought of as simulating (the processes of) a mind rather than being a true one.

Question 22

What is strong AI?

Answer 22

Artificial intelligence that possesses a mind capable of thought and understanding of the area/s it is applied to. Actually possesses a mind and thought - not just simulating a mind.

Question 23

What are some similarities between computers and brains?

Answer 23

???

Question 24

What are some differences between computers and brains?

Answer 24

???

Question 25

What is the determining state of a machine?

Answer 25

The minimal state that describes the elements which drive or determine the behaviour at any stage - so that there is only one possibility for what happens next. For example, the determining state of a binary digital computer may be described as the bit pattern containing all the current instruction(s) and associated data present at the time.

Question 26

True/false: For any particular binary pattern of firing states in a brain there is only one possible next binary pattern. [A neuron firing or not firing may be interpreted as 1 or 0.]

Answer 26

???

Question 27

True/false: Hence there is a determining state for a brain which is its binary firing pattern.

Answer 27

???

Question 28

True/false: In the sense of having the same type of determining state, a brain is therefore equivalent to a binary digital computer.

Answer 28

???

Question 29

True/false: Hence it should be possible for us to build computer programs that rival or exceed humans in their intelligence.

Answer 29

???

Question 30

What is The Chinese room argument?

Answer 30

The Chinese room argument holds that a program cannot give a computer a “mind”, “understanding” or “consciousness”, regardless of how intelligently or human-like the program may make the computer behave.

Suppose that artificial intelligence research has succeeded in constructing a computer that behaves as if it understands Chinese. It takes Chinese characters as input and, by following the instructions of a computer program, produces other Chinese characters, which it presents as output. Suppose, says Searle, that this computer performs its task so convincingly that it comfortably passes the Turing test: it convinces a human Chinese speaker that the program is itself a live Chinese speaker. To all of the questions that the person asks, it makes appropriate responses, such that any Chinese speaker would be convinced that they are talking to another Chinese-speaking human being.

The question Searle wants to answer is this: does the machine literally “understand” Chinese? Or is it merely simulating the ability to understand Chinese? Searle calls the first position “strong AI” and the latter “weak AI”.

Question 31

Can an English-speaking user in the Chinese room gain an understanding of Chinese if also given the following programs? Why?

• a Chinese-English (and vice-versa) dictionary program so that the Chinese questions are answered through the English Speaker’s mind.
• a program that enables questions in English to be
  answered convincingly through a Symbolic AI program.
• a program that carries out the digital logic entailed by
  the symbolic computation of the 2nd modification via a neural network.

Answer 31

Yes (?) ???