The Turing Test

Question 1

What is the imitation game?

Answer 1

Three people- man, woman and an interrogator.

Man can lie to persuade interrogator that they are a woman. Woman tells the truth.

Interrogator has to decide which is which

Now replace man with computer- will the interrogator decide wrongly as often as when playing with real man?

Question 2

Why us it suitable to use the imitation game in place of the question ‘whether computers can think’?

Answer 2

The game tests the things we consider thinking (talking, writing poetry, maths etc.) while not testing things we would want out of scope (physical strength, beauty, etc.)

Question 3

Which machines are concerned in the game?

Answer 3

Original idea is to include any possible technique for creating a machine. But then we could include human clones which we don’t want to do.

Not any kind of machine can be considered, only digital computers (this is only a problem if it turns out that computers cannot win the game, then would need to reconsider the definition of machine in order to be able to answer the question whether machines can think). The question is not whether all computers would do well at the game but if an imaginable computer which does well could exist.

Question 4

What is a digital computer?

Answer 4

• A digital computer can be regarded as consisting of three parts:
  1: Store
  2: Executive Unit
  3: Control
• The store part stores information such as a human could do calculations in his head or paper where he would make notes, store info.
• The executive unit carries out various individual operations involved in calculations
• Control has to follow the table of instructions (which is part of the store- programmed into the system )
• The information in the store will be broken up into packets of moderate small size such as ten decimal digits. Such as “add two numbers”
• The control will obey the order of the positions in which they are stored (can include loops).
• Earlier mentioned instruction tables such as in the human computer cannot only be encoded using programming into a computer.
• An earlier described machine such as Babbage’s Analytical Engine was to be mechanical, disproving that even though the human mind runs on electrical currents a machine such that it could pass the Turing test would not have to.

Question 5

What is the universality of digital computers?

Answer 5

• Digital computers can be considered discrete state machines.
• Technically every machine has intermediary states (even light switches) so there are no discrete state machines but that is unimportant for our result.
• From one state of the machine and the input of the machine all future states can be predicted.
• Laplace view expands this to the entire universe, if we can define the current state of the universe we can predict all future states (of course that’s not really how it works but we can assume so for digital computers)
• The number of states as described in digital machines is enormously large
• Hypothetical experiment with 10^150.000 states combining the number of states in three Manchester machines.
• The special property of digital computers is that they can mimic any discrete state machine (they are universal machines)
• Therefore if any machine can handle it then digital computers can cause they can mimic any other machine
• Can Machines Think” -> ‘Are there discrete state machines which would do well in the imitation game”- can check with digital computers

Question 6

What does Turing think will happen?

Answer 6

• In the year 2000 an average interrogator will not have more than 70% chance of making the right decision after 5 minutes of questioning.

Question 7

What is the theological objection? and it’s counterargument?

Answer 7

OBJECTION: God has only given a soul to humans hence no animal nor machine can actually think.
COUNTER-ARGUMENT:
* God may administer a soul to animal if he sees fit, though he would have to alter the physique of said animal to accommodate. Why could he not do the same for a computer?
* Biblical arguments do not support reality (i.e. previously bible used to refute idea we moved round sun- Copernicus- but now we know better)

Question 8

What is the heads in the sand objection?

Answer 8

OBJECTION: The consequences of thinking machines would be too dreadful lets hope it doesn’t get to this. Because it removes our superiority as humans
COUNTER-ARGUMENT: Argument so weak it does not need refute

Question 9

What is the mathematical objection?

Answer 9

OBJECTION:
* Discrete machines are not limit-less which is a mathematically proven fact
* Gödels theorem proves that logical system statements can be formulated which can neither be proven nor disproven.
* Consider a machine which answers yes or no questions and considers the following question: ‘Consider the machine specified as follows … Will this machine ever answer Yes to any question?”.
* The dots are to be replaced by any description of some machine in a standard form.
* When the machine has any kind of relation to the machine which is under interrogation it can be shown that the answer is either wrong or not forthcoming (because it is a self-reference and the computer gets stuck)
* The argument then follows that there are limits to the powers of any particular machine, no such limits apply to the human intellect (though this is a statement without proof)
COUNTER-ARGUMENT:
o The argument states that there are men that are intellectually superior to a certain machine however we cannot definitively prove that for all machines. There might be machines more clever than all men. We make mistakes often too

Question 10

What is the argument from consciousness?

Answer 10

OBJECTION:
* Argument well expressed in Lister oration
* “Not until a machine can write a sonnet or compose a concerto because of thoughts and emotions felt, and not by the chance fall of symbols, could we agree that machine equals brain.”
* The most extremist view is that only if one can become the machine and feel what the machine is feeling we can discern intellect.
* Solopsist views: Only I know I think therefore I cannot be certain anyone else is capable of thought.
COUNTER-ARGUMENT: All objectivity stemming from this argument would be either extreme solipsist or forced to accept its objection

Question 11

What are arguments from various disabilities?

Answer 11

OBJECTION:
Machines can do everything mentioned so far but not be:
‘’Kind, resourceful, beautiful, friendly, tell right from wrong, make mistakes, think about themselves etc.”
Cannot be intelligent because they cannot do one particular thing

COUNTER-ARGUMENT:
* No support is offered for these statements, seem to be from scientific induction.
* The space-time we have so far explored is by and far not large enough to support scientific induction

‘cannot make mistakes’
* The argument that machines cannot make mistakes can be morphed into the criticism that one may discern a computer using this logic.
* However, the machine would deliberately make mistakes in order to trick the examiner.
* If we are going to argue about mistakes then need to distinguish between mistakes as ”errors of functioning” (mechanical or electrical fault) and “errors of conclusion” (mistakes made during interpretation, i.e. computer that just type 1==0 is following what it was told to do but this is a mistake)
* So the abstract computers we are arguing about cannot make mistakes as in errors of functioning but can with regards to errors of conclusion

‘think about themselves’
* First of all have to establish that computers can think about certain subjects
* But then we know that computers can be used to write programmes for themselves, predict effect of these programmes etc.

• Any criticism that computers cannot have much diversity of behaviour is just a way of saying they cannot have much storage capacity.
• Ultimately think is mainly the ‘argument for consciousness’ disguised. We can describe how a computer might do all of these but how can we really know without being the computer

(Also this argument is very focussed- perhaps incorrectly- on one thing- we can create a Turing like test for any sort of behaviour)

Question 12

What is Lady Lovelace’s Objection?

Answer 12

OBJECTION: Machines can only do whatever we know how to order it to perform. -> they have no pretense to originate anything
COUNTER-ARGUMENT:
* Machines cannot takes us by surprise -> retorted by saying machines do take us by surprise (for example when we make a mistake in the code)
* The assumption that as soon as a fact is presented all possible consequences spring into mind simultaneously. But this is simply false.

(Additionally, computers are learning how to do things without us understanding how they’re doing it more and more- Lovelace’s point seems to no longer apply as much)

Question 13

What is the argument from continuity in the nervous system?

Answer 13

OBJECTION: The nervous system is not a discrete state machine.
COUNTER-ARGUMENT: However with the game in place as-is the examiner should not gain any advantage from the difference as the computer should still be able to do a good impression of ‘more continuous’ systems

Question 14

What is the argument from informality of behaviour?

Answer 14

OBJECTION:
* It is not possible to produce a set of rules purporting to describe what a man should do in every conceivable set of circumstances.
* From this can be argued we cannot be machines.
* “If each man had a definite set of rules of conduct by which he regulated his life he would be no better than a machine. However there are no such rules, so men cannot be machines”
COUNTER-ARGUMENT:
* Humans do live by certain laws of behaviour (i.e. we squeak when pinched)- are we therefore machines?
* The only way we can be convinced of the absence of complete laws of behaviour as a complete set of rules of conduct is by scientific observation.
* There are no circumstances in which we can definitively say we have searched enough there are no such laws.
* However with the many states a discrete state machine has it is also impossible to predict what a new input to a machine will yield says Turing comparing man to machine.

Question 15

What is the argument from extra-sensory perception?

Answer 15

OBJECTION:
* The notion of telepathy, clairvoyance, precognition and psycho-kinesis. These phenomena seem to deny all our scientific ideas.
* Telepathy has statistical evidence.
* One can claim that many scientific theories remain workable with E.S.P
* Suppose one of the players has E.S.P and can guess a card more often than random.
* Perhaps psycho-kinesis might cause the machine to guess right more often than would be expected on probability calculation.
COUNTER-ARGUMENT:
* If telepathy is admitted the test needs to be tightened to put the competitors into a ‘telepathy-proof room’

Question 16

What can we therefore conclude about learning machines?

Answer 16

• No very convincing arguments for the test itself only good refutations for objections.

Two sort of arguments?
* Lady Lovelace argued that machines can only perform what we explicitly instruct them to do. Turing counters by suggesting that a person can “inject” an idea into a machine, causing a response that might eventually fade away. He uses the analogy of a piano string struck by a hammer or an atomic pile. If the machine is made larger (analogous to a critical-size pile), the disturbance caused by an idea could potentially escalate without bounds. Turing questions whether similar phenomena exist for both human minds and machines, exploring the idea of machines becoming “super-critical” and generating extensive responses.
* Turing introduces the “skin of an onion” analogy to explore the layers of functions in the mind or brain. Initially, some functions can be explained mechanically, but Turing suggests these are like outer layers that need to be stripped away to reach the “real” mind. The question arises: does this process lead to a genuinely non-mechanical core, or does it eventually reveal a layer devoid of substance? Turing raises the possibility that the entire mind might be mechanical but emphasizes that, even if it is, it would not necessarily conform to the discrete-state machine model previously discussed.

• Advances in this problem should stem from advances in programming rather than engineering.
• Imitating an adult human mind three components:
  o The initial state of the mind say at birth.
  o The education to which it has been subjected
  o Other experiences not described as education
• Presumably the child’s mind is more suitable for learning purposes.
• We cannot expect to find a good child-machine at the first attempt one must experiment with teaching one such machine and see how well it learns.
• Comparing to evolution- natural selection, hereditary material, mutations
• Normally punishment and rewards are associated with the teaching process.
• It is necessary to have some unemotional channels of communication/
• It is possible to teach a machine by punishments and rewards to obey orders given in some language e.g. a symbolic language.
• The processes of inference used by the machine need not be such as would satisfy the most exacting logicians.
• The idea of a learning machine might seem paradoxical. How do rules of the machine change.
• The rules that change are of a rather less pretentious kind, claiming only an ephemeral validity.
• The teacher of a learning machine will often be largely ignorant of what is happening within the machine.
• This challenges the view that machines can only do what we know how to order it to do
• Processes that are learnt do not produce a hundred per cent certainty of result; if they would they could not be unlearnt.
• Better to have some randomness in finding optimal behaviour (might be large chunk of bad behaviour so don’t want to go sequentially)