Descartes creates two rules to distinguish machine thinking from human thinking. The first rule is that machines ‘could never use words, or put together signs, as we do in order to declare our thoughts’ (Study Guide:105). Although he states in Discourse on a Method that machines can ‘[utter] words’ (Study Guide:105), they would never be able to do it in the manner as we do. Their responses would lack ‘appropriately meaningful [answers]’ (Study Guide:105) if they deviated from the arrangements programmed into them. No matter how effectively machines mimic human behaviour, only an immaterial thinking substance (our mind) could participate in the flexible use of language by responding appropriately to unforeseen circumstances. Human speech is a non-mechanical action, and no matter how complex the programming of a machine, their replies would be directed by rules.
The second rule is focused on reason and common sense. Even though machines can behave the same way as humans do, they would not be acting from ‘understanding but only from the disposition of their organs’ whereas reason is our ‘universal instrument’ (Study Guide:105). This second method suggests that they would be mimicking human behaviour in a restricted number of ways; there is not enough space for organs to fit inside a machine to give it the ability to reason through every unpredictable event the machine was to encounter. People, on the other hand, can partake in conversations on topics they know nothing about, but still create meaningful replies. Hence, it is impossible for machines to act to the same capability that reason allows us to.
Fast forward to the 20th century and the Turing Test challenges Descartes’ views. First of all, I will examine his functionalism definition of ‘thinking’. The mind is akin to a computer program with inputs (i.e. sensory information, such as touching a hot pan) and outputs (i.e. reactions, such as moving the hand away from the pan in pain). Thinking is just a computational process (Study Guide:112). This contrasts to Descartes as thinking is defined by consciousness and an immaterial mind.
According to Turing, the test is an indicator of artificial intelligence. The human interrogator asks a computer a series of question. Through their conversation, the interrogator is to determine whether or not the other is a machine or not. If the machine’s behaviour is indistinguishable from a human and the interrogator is successfully deceived, then the computer is thinking and intelligent (Turing:128).
I will now evaluate both arguments. The Turing test is less rigorous than Descartes’ two rules. If Descartes’ obstacle to producing a reasoning machine is its inability to use language flexibly, then a machine that passes the Turing Test is sufficient in proving an object truly thinks. As long as a computer can deceive the interrogator into thinking it is a person, it passes the test. Simply passing is sufficient for thought. In contrast, the Cartesian test considers more factors than the Turing test. It tests both the machine’s language abilities and the ability to solve problems in everyday situations. Hence, I think that a machine could pass the Turing test without passing Descartes’ test by failing to use reason. The Cartesian test is then a more appropriate goal for artificial intelligence.
Descartes’ argument, however, is not without problems. It stumbles in the light of technological advancement. Firstly, it raises a few questions concerning Cartesian dualism. As stated before, thinking is defined by consciousness. Theoretically, if we were to fully understand the mental realm and possess the technology, we could possibility manipulate it as we do the physical. If there is a causal interaction between the physical and mental realms (Study Guide:126), we might shape the mental substance in a manner sufficient for reason. Although we do not know how to achieve this and is seemingly far-fetched, if the immaterial can be created and manipulated, then thinking machines can be possible in Descartes’ view.
Secondly, Descartes states that machines cannot act meaningfully in unforeseen circumstances. It is not entirely impossible for a machine to be created where its algorithms give it the capability to create meaningful conversation in unpredicted circumstances, paralleling the human experience of learning. For example, chatbot programs like Cleverbot ‘learns’ from its conversations with people. It stores information in a database and mimics past human responses in future conversations.
However, despite Descartes’ argument weakened in the face of technological advancement, the crux of his argument should not be invalidated. His rules to test the meaningful use of language and reasoning abilities are more appealing to our intuition than Turing’s. In the Turing test, there is no need for an immaterial thinking substance. The test is rather clinical, disregarding qualities like emotions or motivation that sets us apart from machines. These qualities, the act of thinking about something is known as intentionality, and is discussed by Searle. I will present his argument below.
Searle’s Chinese Room offers a different perspective. The thought experiment is a scenario where a person in a room receives papers with Chinese on them. He also receives an English instruction book that contains rules, aiding the person to match a set of characters to another to from meaningful responses. When his replies are taken out of the room, they are no different from native Chinese speakers (Searle:144). The process parallels computer process, where Turing’s machine has inputs and outputs and symbol matching. There is no difference between the role of the computer and the person. Each follows a program, and mimics behaviour.
This experiment does aid the debate by aiming to show that artificial intelligence is false and that the Turing Test is not adequate. It also appeals to our intuition. The person in the room imitates the syntax (the grammar, such as sentence structures) of a native speaker, but lacks semantic understanding – failing to understanding the actual meaning of Chinese (Study Guide:115). They also lack intentionality – failing to think about Chinese in virtue of the task. Similarly, the computer relies on syntax, and cannot interpret the language solely based on semantics as native speakers do. This symbol manipulation cannot equate to intentionality and understanding. Therefore, the Turing Test does not show true thinking, and only demonstrates a mimicked behaviour.
Furthermore, as Descartes and Searle suggests, true understanding requires more than symbol manipulation. For example, internet memes have recently gained popularity in everyday life. A meme is usually a phrase that communicates social and cultural ideas. ‘Damn Daniel’ was a phrase that relayed solidarity and positivity when commenting on a person’s outfit. It also contained humour and was considered an inside joke. This social semantic reaches beyond dictionary definitions. Based on pure symbol manipulation, a machine cannot process the humour and implied social semantics attached to the phrase and create a meaningful response to it. Hence artificial intelligence cannot truly think.
Turing’s test focuses on imitation, but Descartes’ rules are more rigorous. Aided by Searle, the argument for semantic understanding overrules Turing’s views that machines can think.
Bibliography
Faculty of Arts (2014), Time, Self and Mind Study Guide. Melbourne, Monash University.
Turing, Alan (1950) ‘Com