The debate on whether inanimate objects can or cannot hold any type of conscious state has been around in the modern world for a while now. Rene Descartes is amongst scholars who have questioned the capability of non-animals to hold reasoning faculties same as the ones for humans. To prove that inanimate objects are not capable of holding a conscious state, Descartes came up with two tests which include; language use and universal problem-solving. In terms of language use, he maintains that inanimate objects such as robots are not able to come up with original sentences or words that no other human being has thought of. He also opines that inanimate objects have no ability to reason and have independent thinking. However, modern technology has brought about inanimate objects such as chatting bots that are able to talk. Additionally, there have arisen other tests that check intelligence such as the Chinese Room Argument, Turing, and Lovelace. This essay aims at discussing whether it is completely true that language use and universal problem solving are the only tests that can be used to check human intelligence and whether inanimate objects are the only ones that have an unlimited intelligence and consciousness.
Machine intelligence has developed over time with many intelligence tests emerging to prove that machines are capable of possessing a conscious state. The 1950 Turing Test that was developed by Alan Turing was able to trick humans into believing that it is indeed a human. Turing argued that because it is able to trick humans in such a way then it has intelligence (Marsden, 2017). In 1955, John McCarthy coined the word artificial intelligence to describe how intelligent machines can be made scientifically. Then in 1961, Unimate which was the first industrial robot began working at GM taking the place of human workers on the assembly line (Marsden, 2017). The chatbot known as Eliza was developed in 1964 by Joseph Weizenbaum and it had the capability of holding conversations with humans. All these non-human machines were made in an attempt to reveal that intelligence is not limited to humans only but inanimate objects. Moreover, in 2014, Amazon launched Alexa, an intelligent virtual assistant who was capable of performing shopping tasks (Marsden, 2017). However, one major difference between these inanimate objects and humans is that whereas they are limited to certain tasks, humans’ capabilities are unlimited. Humans are able to use words and other signs to declare their thoughts, unlike inanimate objects which cannot put together words on their own. Although the inanimate objects such as the Unimate can perform several tasks effectively even than humans, they cannot function in tasks other than the ones that they are programmed for (The Ideal Lab, 2013). Therefore, it is correct to say that these objects are not acting because of their own knowledge but due to the disposition of their organs.
The Turing Test is a major thing in machine learning because Alan Turing, the creator came up with it to reveal that computers are capable of having brains just as humans do. In 2014, the Turing Test competition was held (Mann, 2014). Several celebrity judges were brought on board to try to prove that computers are intelligent. A big number of programmers took part in the competition and the winner of the competition was Robert Llewellyn who had created a program known as Eugene Goostman. Goostman was able to convince a third of the 30 judges that it was indeed a real boy. Llewellyn had programmed Goostman that it was able to clearly state that it is 13 years old and that it is a Ukrainian non-native English speaker (Mann, 2014). Also, it managed to trick the judges in 30% of its presentation. However, the capabilities of Goostman cannot be equated to humans. Goostman was programmed to give certain specific responses to questions but was not able to generate answers to questions that it had not been programmed to give. Humans are able to process information and reason without having a fixed way of response. Their thinking is versatile because their brains are wired such that they can make sense of things that are not familiar to them by comparing them with their existing knowledge (The Ideal Lab, 2013). This is something that Goostman lacked and therefore, Descartes’ argument that language is the major test of human intelligence is correct.
After the Eugene Goostman chatbot was said to pass the Turing’s measure of machine intelligence, it became apparent after an hour later that the celebration was not worth as it had been thought. In fact, the chatbot proved that it had very little knowledge compared to human intelligence. Selmer Bringsjord who is one of the people who designed Lovelace intelligence test was very critical about the Turing Test which he described as inadequate (Pearson, 2014). The Turing Test works by pitting a human interlocutor against a computer program and its main trick is to make humans think that it is human so as to prove that its intelligence matches humans’ intelligence. The main trick is to come up with a program that can make a person believe that he or she is interacting with a human. In the case of Goostman, by feeding the chatbot with the age and foreign nationality information made it win over ten out of thirty judges thus passing the Turing Test. Bringsjord downplays this method because it is based on mimicking human’s basic language skills and not demonstrating genuine intelligence in a machine. The chatbot is able to order words, but it is incapable of figuring out their meaning (Pearson, 2014). He goes on to argue that the machine can only be said to be intelligent as a human if it proves that it can originate with an idea that it was not designed come up with. Otherwise, it is only a mimicking machine and not an intelligent one. These arguments support Descartes’ argument about language being the primary measure of intelligence.
Bringsjord considered his program, Lovelace Test better than the Turing Test because it tests genuine autonomous intelligence. This tool does not only look for the capability of an intelligent machine to sound like a real human but also tests human-like creativity and origination (Pearson, 2014). It does not simply manipulate syntax. The intelligence machine must be able to output a new idea that its designer must not have the capability to explain where it originated from. It tests true machine cognition and not a replication of what the tool was programmed to perform. Due to this, Bringsjord believes that it is impossible for any intelligence machine to pass the Lovelace Test. It is impossible for a programmed machine to have the free will to self-determine things because it doesn’t have an autonomous intelligence (Pearson, 2014). Moreover, even the self-learning neural network that is considered the most advanced is only capable of performing tasks that are mathematized and coded. According to Pearson (2014), only humans have the ability to have social cognition and it is impossible for this to be mathematically formalized. Therefore, Descartes’ argument that human intelligence can only be measured through language use and universal problem solving is still correct.
According to Oppy and Dowe (2016), it is possible for machines to imitate human actions as closely as possible but it is still very possible to note that they are not real men. One criterion that they propose is the use of words coupled with signs. Machines do not have the capability to produce different word arrangements so as to respond to a particular question. The dullest men are able to appropriately offer meaningful responses to questions posed to them because their brains are able to play with different words based on the context. The second criteria that Oppy and Dowe (2016) propose is that although machines can function well by performing tasks that are meant for humans, they cannot succeed in other functions that they are not programmed to perform. Machines only act based on the disposition of their organs and therefore it is impossible for them to perform different functions because they cannot bear several different organs. However, humans have the capability to perform and learn different tasks because they have universal problem-solving capabilities.
Hanson (2012) opines that human language is so contextual and so ambiguous such that it is impossible for a machine to perform in a similar manner. Machines are fed with data that is specific for a particular task and context and if they are supposed to respond to a different context they will not be able to succeed because it is beyond their capability. The IBM Watson may pass the Turing test but is not capable of being compared to human intelligence which is diverse (Hanson, 2012). Therefore, language remains to be the major measure of human intelligence.
Cole (2004) uses the Chinese Room Argument to discredit the Turing Test. In the Chinese Room Argument, Searle uses a computer program to manipulate Chinese numerals and symbols to an extent that he fools the people in the room that he understands Chinese. The program might appear to understand the Chinese language because it produces strings of Chinese characters that seem real, but in the real sense there is no real understanding (Cole, 2004). The Turing Test works in a similar manner and therefore cannot be said to be an adequate tool for measuring machine intelligence. The test does not have an understanding of semantics or meaning of whatever it says because it just simulates responses (Cole, 2004). Therefore, it is correct to argue that the Turing Test does not prove that machines possess consciousness as humans do.
Conclusion
Although machine intelligence has advanced over years to make machines function even better than humans, there are two aspects that set humans aside in terms of their capabilities. Descartes opines that language use and universal problem solving are the only measures that can be used to test human intelligence. This argument is true because, despite machines being purported to possess the ability to communicate effectively with humans, their capabilities are still limited. The Turing Test was set up to measure human intelligence but it has failed miserably because some of the machines that have performed well in it prove otherwise. It is possible for intelligent machines to offer responses to questions that it is programmed to respond to but if placed in another context it cannot provide adequate answers as humans can. Another measure of human intelligence is the universal problem solving whereby humans are able to perform different tasks with the same organs whereas intelligent machines cannot. Therefore, Descartes’ argument concerning human intelligence and machine intelligence still hold water.
References
The Ideal Lab. (2013). The Idea Lab: Philosophy of Artificial Intelligence from a ... Retrieved from http://brandonhtomlin.blogspot.com/2013/05/the-applicability-of-rene-descartes.html
Cole, D. (2004, March 19). The Chinese Room Argument. Retrieved December 19, 2018, from https://plato.stanford.edu/entries/chinese-room/
Hanson, D. (2012, April 18). Can IBM Watson pass the Turing Test? Retrieved from https://www.youtube.com/watch?v=bfLdgDYjC6A
Pearson, J. (2014, July 08). Forget Turing, the Lovelace Test Has a Better Shot at ... Retrieved December 19, 2018, from https://motherboard.vice.com/en_us/article/pgaany/forget-turing-the-lovelace-test-has-a-better-shot-at-spotting-ai
Mann, A. (2014, September 06). That Computer Actually Got an F on the Turing Test | WIRED. Retrieved December 19, 2018, from https://www.wired.com/2014/06/turing-test-not-so-fast/
Oppy, G., " Dowe, D. (2016, February 08). The Turing Test. Retrieved December 19, 2018, from https://plato.stanford.edu/entries/turing-test/
Marsden, P. (2017, August 21). Artificial Intelligence Timeline Infographic – From Eliza ... Retrieved December 19, 2018, from https://digitalwellbeing.org/artificial-intelligence-timeline-infographic-from-eliza-to-tay-and-beyond/
