The Turing Test and Beyond Flashcards
Explain what is meant by a criterion. Specify two types of (logical) criterion that were discussed in the lectures.
A necessary criterion for the attribution of intelligence explains that everything that is intelligent is P:
S is intelligent -> S possesses property P
A sufficient criterion for the attribution of intelligence explains that everything that is P is intelligent:
S possesses property P -> S is intelligent.
What is meant with biological criteria for the attribution of intelligence, name two examples and provide two challenges to this criteria.
Biological criteria stipulate that the properties relevant to the attribution of intelligence are the properties of biological organisms.
Examples include:
- Constitution: S is made of organic matter (flesh)
- Function: S plays a role in, and is adapted to, a biological ecosystem
(other examples are also possible)
Two challenges include:
- Biological criteria are overly chauvinistic. Chauvinism is the unreasonable belief in the superiority or dominance of one’s own group or people. Non biological entities are excluded from consideration from the start. Is this justified?
- “Excessive liberalism” refers to the idea that the criteria for intelligence are too broad and inclusive, such that they would include organisms that many people would not traditionally consider to be intelligent, such as plants
What is meant with computational criteria for the attribution of intelligence, name two examples and provide two challenges to this criteria.
Computational criteria stipulate that the properties relevant for the attribution of intelligence are the ones that define certain classes of computational systems.
Examples:
- S is a PSS
- S is a Turing machine
Challenge:
- Challenge of computational limitations: Does intelligence have features that cannot be replicated computationally
- Challenge of cognitive scientific uncertainty: We don’t even know (yet) that human beings really are computational systems, so why suppose that any kind of computation is sufficient for intelligence?
What is meant with behavioral criteria for the attribution of intelligence, name two examples.
Behavioral criteria stipulate that the relevant properties for the attribution of intelligence are a system’s (overt and measurable) behavioral properties.
Examples:
- S passes the IQ test
- S wins at popular games such as chess
Give a concise description of the Turing test, also explain why it is an example for a behavioral criterion.
The test involves a human evaluator engaging in natural language conversations with both a human and a machine, and determining which of the two they are conversing with. If the evaluator is unable to distinguish the machine’s responses from those of a human, the machine is said to have passed the Turing test and demonstrated human-like intelligence.
This is an example of a behavioral criterion, because it is based on the observable behavior of the machine, specifically its ability to engage in natural language conversation in a way that is indistinguishable from a human. It does not rely on any assessment of the internal workings of the machine or the specific mechanisms it uses to generate its responses.
Provide four objections to the Turing Test.
The heads in the sand objection: The consequences of machines thinking would be too dreadful.
The argument from consciousness: The Turing Test only focuses on a machine’s ability to mimic human behavior, but it fails to take into account the subjective experience of consciousness. This argument posits that true intelligence must include consciousness, and that a machine that is able to mimic human behavior without actually experiencing consciousness cannot be truly intelligent.
The argument from various disabilities: The imitation game as only considers verbal behavior. But this is an arbitrary restriction. Analogous scenarios could be designed to consider many other kinds of behavior.
Lady Lovelace’s objection: It is not the machine that is intelligent, but the behavior of the machine reveal more about the intelligence of a machine’s programmers.
What is meant with the Total Turing Test criterion?
Perhaps it is not sufficient for S to possess any one of these
behavioral capacities. Perhaps it must possess them all (at the
same time).
S “fools” humans in any conceivable context → S is intelligent
Which objection to the Turing Test becomes less appealing when you consider ‘learning’ machines.
Lady Lovelace’s objection. The programmer does not explicitly program the rules into the machine, but is only responsible for the learning algorithm and the data it can learn from. Therefore, our ability to control and predict the machine’s behavior decreases, which increases the gap between programmer and machine.
Explain the “Threat of Pretence”.
Can something behave “as if” it is intelligence, without actually being intelligent? It goes to the very core of our intuitions about what it means to be intelligent: does it merely depend on what you are capable of doing, or also on how you do it?
Note that the assumption that the possibility of pretense is philosophically significant is not so much an argument against the claim that behavioral criteria are sufficient, as much as it is a denial thereof!
