Software developers and philosophers often disagree about whether a computer can have an individual mind similar to that of humans. Programs called artificial intelligence are installed in many modern devices and are part of everyday life. However, not everything is clear about this issue due to several counter-arguments. The world-famous experiment, called the Chinese Room, shows that, even with the ability to process information correctly, the computer follows the given algorithm without showing real cognitive properties and only following the exact path. To refute the idea that programs have real intelligence, the connectionist approach will be used, and the answer will be largely based on the idea of deep learning. The Chinese Room Argument is an example of the fact that an artificial neural network, in spite of demonstrating complex calculations and inferences in analysis, cannot process information at the level of human consciousness.
Summary
The Chinese Room Argument is a globally renowned experiment that was designed to disprove the idea that computers’ artificial intelligence can be compared to that of a human. It was carried out by John Searle in 1980, and today, many sociologists and philosophers are studying it (Alberts, 2020). The evaluation algorithm involved the use of symbols, namely Chinese characters, which a person had to use to answer pre-prepared questions. According to Agar (2019), the main reason for the experimenter’s doubts was that, being ignorant of the meaning of the proposed characters, a person could only apply them through a pre-given algorithm. As a result, the conclusion was developed that the principle of operation of computer programs with a learning option was similar. Machines can only execute a certain set of commands by following a specific path, thus simulating conscious activities. In reality, one cannot talk about intelligence or rational thoughts such programs can demonstrate.
Such reasoning, based on the inadmissibility of the presence of real intelligence in a computer, proves that even the physical manifestations of the mind cannot prove the existence of a fundamental mind. Software needs to go through the full cycle of perceiving, understanding, and interpreting specific information to be a thinking entity. Chinese characters, being unfamiliar symbols for a person, cannot be interpreted adequately if, initially, a person does not have knowledge. The combination of symbols cannot be randomly processed for the responses to be received to be adequate and semantically consistent with the questions posed (Alberts, 2020). Therefore, from any perspective, the idea of such an experiment refutes the ability of a computer to exhibit a thought process that is available only to a person.
The fact that a computer is capable of exhibiting signs of intelligence is not comparable to the idea that it is inherently intelligent. Based on the details of the experiment, Louwerse (2018) notes that symbol manipulation is nothing more than a consequence of a programmed binary code. In other words, from a philosophical perspective, such knowledge is limited and based on prior training. A computer program can be considered one that does not have the flexibility of perception and interpretation. With the same success, the program could work with other characters and texts. The idea that machines are capable of self-learning runs counter to the theory of symbolism that characterizes the human brain (Lyre, 2020). The perception of images and signs can be intuitive, and with learning and analysis, a person is able to acquire new skills, which the computer cannot do, as the result of the experiment proves. Thus, despite the external similarity of learning, human and computer methods differ significantly.
To address the proposed issue from a philosophical perspective, the connectionist/deep learning approach can be used as a background for interpretation. By adapting relevant ideas to the experiment in question, its validity can be assessed in terms of rationality and objectivity. A direct connection between the work of a computer program and the functioning of the human brain should be made to identify the prerequisites for the manifestation of the mind, as well as prospects for learning. Along with the key idea, possible counter-arguments can be presented to determine how reasonable they can be in the context of the topic raised.
Argument
The application of the connectionist/deep learning reply to the argument under consideration about the impossibility of considering artificial intelligence as having consciousness confirms the thesis that such a correlation is impossible. For instance, Louwerse (2018) emphasizes that any neural networks work as an imitation of the functions of the human brain. Searle’s argument about the inadmissibility of comparing a person and a machine is justified because the internal motives of consciousness, namely the ability not only to perceive but also to interpret information flexibly, is a semantic tool of consciousness. Therefore, one cannot say that computer programs can perceive data in the same way as people since comprehension requires the presence of cognitive functions, which is not available to machines.
From the standpoint of connectionism, the human brain can be called an example of a program that has the depth and complexity sufficient to make reasonable judgments. Lyre (2020) remarks that Searle’s view of the experiment suggests the existence of artificial intelligence within a given framework. However, regarding the connectionist theory as a background for analysis, one may observe that the formal symbols, which, in this case, are Chinese characters, are not sufficient to build a mind.
A computer system may include intelligent and reasonable assumptions; consequently, it is nothing more than a consequence of pre-planned evaluation algorithms but not the manifestation of cognitive activity. The complexity of the data, namely the characters, is an obstacle to demonstrating real learning abilities. As a result, connectionism does not deny the possibility of artificial intelligence, but the mind as a property is denied in such systems.
In general, an artificial neural network is a set of embedded steps and algorithms that function through pre-built tactical decisions. Given the concept of deep learning, semantic segmentation is a natural property of a system that is capable of cognitive development (Zhang et al., 2019). With respect to the Chinese Room Argument, the computer deals primarily with syntactic structures that are incompatible with semantics. In other words, artificial intelligence interprets data in accordance with one specific principle, but understanding information, as one of the main properties of the human brain, is not available to it.
The experimenter uses formal symbols as an example of an activity that can reflect the ability to understand. For a computer program, a character set remains a character set and nothing else because the essence of the language embedded in these characters cannot be conveyed if the basic perception of the language is missing. This assumption also speaks in favor of the fact that the concept of deep learning does not allow one to evaluate artificial intelligence as a system capable of actual learning comparable to that of humans.
When speaking about the inconsistency of the connectionism approach, one can point out counter-argumentary ideas. For instance, the perception of artificial intelligence as a system capable of building logical chains brings it closer to the human mind. Moreover, since modern computers have advanced significantly beyond what was available to researchers in the 1980s, the progress in working with data is evident. Liu et al. (2019) give an example of semantic segmentation performed by modern neural networks and note a significant increase in the capabilities of computer systems. Therefore, within the framework of an alternative idea, the mind as a property characterized by the ability to isolate individual pieces of information and distribute data can be considered in relation to advanced computer programs.
At the same time, the considered counter-arguments can only be regarded as partially justified. Artificial intelligence can match and analyze data in logical chains. However, the most important property that could bring it closer to the abilities of the human brain is related to understanding. Computer programs do not have the function of cognitive perception, which does not allow them to be regarded as intelligent. In terms of progress in the field of computer technology, modern digital systems have impressive capabilities. Nevertheless, they still lack the interpretive skills inherent in humans, which rules out the possibility of the mind in artificial intelligence. Therefore, the connectionist/deep learning approach seems reasonable when considering the experiment in question and denying consciousness in computer systems.
References
Agar, N. (2019). How to treat machines that might have minds. Philosophy & Technology, 33(2), 269-282. Web.
Alberts, L. (2020). Not cheating on the Turing Test: Towards grounded language learning in Artificial Intelligence [Unpublished doctoral dissertation]. Stellenbosch University.
Liu, X., Deng, Z., & Yang, Y. (2019). Recent progress in semantic image segmentation. Artificial Intelligence Review, 52(2), 1089-1106. Web.
Louwerse, M. M. (2018). Knowing the meaning of a word by the linguistic and perceptual company it keeps. Topics in Cognitive Science, 10(3), 573-589. Web.
Lyre, H. (2020). The state space of artificial intelligence. Minds and Machines, 30(3), 325-347. Web.
Zhang, L., Lim, C. P., & Han, J. (2019). Complex deep learning and evolutionary computing models in computer vision. Complexity, 2019, 1-2. Web.