Is Artificial Intelligence Actually Intelligent?
Defining intelligence through a philosophical viewpoint.
Artificial Intelligence (AI) is everywhere, whether it is powering our Google searches, enabling voice recognition, or even filling our TikTok feeds. With over 37% of organizations implementing some form of AI and a projected market evaluation of $62 billion by 2022, AI has undoubtedly become a monstrous force.¹ ² However, with this sudden spread of AI and machine learning, it is important to take a step back and ask, what are the limits of AI? Does AI have the capacity to become intelligent enough to surpass humans, or are machines fundamentally incapable of being intelligent? First, let’s answer what is Artificial Intelligence, and how does it compare to Human Intelligence?
What is Intelligence?: The Mind, res cogitans
When we think of intelligence, it is easy to associate it with being smart, conscious, or sentient. But what causes something to act intelligently? One argument is that to be intelligent, one must have a mind. The 17th-century philosopher René Descartes proposed that humans possessed a separate intangible matter which he called res cogitans. Res cogitans literally translates to “thinking thing,” and it is essentially the substance that forms mind and consciousness. Descartes believed that humans were the only ones that possessed this substance and consequently were the only ones capable of thinking.
To justify his claim of thinking as unique to humans, Descartes sets up a simple thought experiment in part five of his treatise Discourse on Method. If one creates an artificial monkey machine with the same organs and mechanical functions as a monkey, there is no way of distinguishing between the fabricated and real monkey. However, if one were to create an artificial human machine with the same organs and mechanical properties as a human, Descartes proposes that there would be two means of distinguishing a real human from an artificial one.³
1) The machine would never be able to communicate proficiently through language
2) The machine would be unable to perform a multitude of different tasks.
Descartes claims the root of these two differences between machine and human lies in the presence of res cogitans. It is what enables the most incapable of humans to describe their thoughts while the smartest animals are unable to form even a single sentence. It is what enables the mute and deaf to say more with their hands than birds can sing with their songs. The applications of res cogitans are also universal; a human who writes a book is also able to do multiplication, whereas a machine may excel at one task, but it cannot excel at them all. A clock can tell the time better than any human, but not achieve more than that because its behavior is restrained to the alignment of its mechanical parts. Descartes’ interpretation of intelligence is that it is an incorporeal substance that allows humans to exceed the limits of their physical bodies.
Descartes’ belief that man-made machines that lack res cogitans are unintelligent, suggests that AI is also incapable of being intelligent, as they too lack the res cogitans necessary. While Descartes’ ideology of an immaterial mind is compelling, it has its flaws. Descartes’ theory suggests that animals are not just less intelligent than humans but that they have no intelligence at all. In his theory, animals are reduced to mindless machines, intrinsically inferior to humans, and subject to abuse. Descartes’ reliance on an otherworldly substance acts as a deus ex machina in this plot of defining intelligence. Ultimately, Descartes’ definition of intelligence collapses because of its foundation on the supernatural.
What is Intelligence?: The Turing Test
We will now look at an alternative definition of intelligence from almost three centuries in the future by Alan Turing, the father of computer science and artificial intelligence. Turing proposes that the cause of intelligence is irrelevant; if something acts intelligently, then it is intelligent. Turing defines intelligence based on effect rather than a specific cause.
In his essay Computing Machinery and Intelligence, Turing designs an experiment to test if a machine is intelligent, which he calls the Imitation Game.⁴ The objective is simple: determine if a machine’s behavior can be distinguished from a human. In this experiment:
There are 3 participants: the human, the machine, and a judge.
The human and machine are randomly assigned a label, “A” or “B”
The judge must decide whether “A” or “B” is the human.
If the judge incorrectly guesses the machine as frequently as the human, then we can conclude that the machine’s behavior is indistinguishable from a human.
In the original experiment, the machine is tested solely on its ability to communicate and answer questions, with external factors such as appearance, voice, and handwriting eliminated. Turing also establishes that this experiment should only consider digital machines or digital computers. The physical characteristics of these digital machines (also known as Turing Machines) should not matter or one could hypothetically create a machine of human cells that passes the experiment.⁴ This result would only establish that a biological clone can think rather than a created machine. Turing’s Imitation Game provides an easily replicable experiment to test for intelligence.
This experiment, now known as the Turing test, provides a second perspective on intelligence. Rather than put forward another source of intelligence, he rejects the notion that intelligence has a direct source. The Turing test acts as both a test and definition for intelligence, where intelligence is defined by capability rather than means. Turing liberates intelligence from being exclusive to humans, paving the way for intelligent machines.
The Importance of the Brain
“What do you think my brain is made for
Is it just a container for the mind?”
-Frank Ocean, Pink Matter
Suppose a computer passes the Turing Test, can we definitively say that it is intelligent? Is the cause for the machine’s behavior irrelevant? John Searle, one of the most prominent critics of AI says “No.” Searle argues against the belief of Strong AI, which is the belief that an “appropriately programmed computer really is a mind” (Searle, 417). Searle believes that while AI acts as a good tool for studying intelligence, AI itself is not intelligent. Searle believes that a purely digital machine, such as a computer, is incapable of being intelligent.
During his clash against Strong AI in the 1980s, Searle conceived a thought experiment in his paper Minds, Brains, and Programs. In this experiment, he imagines that he is in a closed room with nothing but a book. He receives papers with abstract symbols written on them from outside the room. Within the book, Searle finds a set of English instructions of what symbols to return based on the symbols he is given. Following the instructions, he returns papers with the respective symbols to the ones he received.⁵ The name of this thought experiment is the Chinese Room argument, and the “abstract symbols” that Searle has been receiving are in reality Chinese characters. Searle has unknowingly been receiving questions in Chinese and returning the correct Chinese answers. Searle argues that even though he appears to understand Chinese, the Chinese characters are completely arbitrary and meaningless to him, just random lines and scribbles on a piece of paper. He believes this is exactly how computers operate. Just as Searle, who blindly follows the book, does not understand what he is doing, a CPU that runs a program does not understand its function. A computer that is given a program that answers Chinese questions does not necessarily understand Chinese. Similarly, a computer programmed to act intelligently is not necessarily intelligent. Searle contends that Strong AI falls into the same dualist fallacies as Descartes. By defining intelligence through a computer’s program, they have once again severed intelligence from the physical.
While Searle asserts that computers are unable to be intelligent based on their program alone, he did not claim that Artificial Intelligence as a whole is impossible. Searle believed that machines could be intelligent, but they needed the right programs and the necessary physical properties. What computers lacked were the physical properties of the brain, specifically intentionality. In Margaret Boden’s book AI, she writes about how Searle maintained that “Intentionality, is caused by neuroprotein much as photosynthesis is caused by chlorophyll. Neuroprotein may not be the only substance in the universe that can support intentionality and consciousness. But metal and silicon, he said, obviously can’t” ⁶. Searle believed that intentionality was an intrinsic property of neuroproteins that allowed humans to give meaning to symbols.⁶ Searle believes that to be intelligent, one must have both syntax, the rules governing words, and semantics, the meaning behind words, the latter of which can only arise from intentionality. Searle suggests that because Turing only tests machines on their syntax but not the semantics the Turing test falsely identifies computers that lack meaning as intelligent. For Searle, a truly intelligent machine needs to have a purpose behind its actions, which can only arise from the physical property of intentionality in neuroproteins.
Are neuroproteins the secret to unlocking intelligence? Or is Searle’s argument inherently flawed? Who is to say that intentionality needs to be a physical property and not just a series of connections and their strengths that can be mapped digitally? When Searle attributes intelligence to a physical characteristic only found in neuroproteins, he ironically falls into a comparable fallacy to that of Descartes. In an attempt to distinguish human intelligence, he attributes all intelligence to an inexplicable property of intentionality found in the human brain and its neuroproteins.
Intelligence: A Spectrum
The definition of intelligence is elusive and seemingly undefinable. Descartes needed a mystical matter to explain it. Searle, who rejected Descartes’ dualism, ironically establishes an equally ambiguous property of intentionality in order to explain the meaning needed behind intelligence. Even Turing’s test has its issues. Animals and babies are unable to pass the Turing test through communication but we can’t say that either of them are unintelligent. What if this is because the question of intelligence is flawed? We should not be asking if something is intelligent, rather we should be asking how intelligent it is.
I propose a redefinition of intelligence and the way we look at it. There is no definite absence or presence of intelligence, rather intelligence exists as a gradient. We should ask how an animal’s or machine’s intelligence compares to humans. I also suggest that there is no absolute universal intelligence; the term intelligence is an umbrella term that embodies an innumerable number of specific instances of intelligence. Intelligence can apply to any capability, and can also be infinitely divided into more specific capabilities. A person could be very mathematically intelligent, but at the same time not very musically intelligent. Their “mathematical intelligence” also acts as an umbrella term, a general measure of one’s ability in all the fields of math that can be split into fields such as: “algebraic intelligence,” “geometric intelligence,” “topological intelligence.” Intelligence is not a simple observable trait, it is a multidimensional spectrum that measures capability.
In their search for a definition of intelligence, Descartes and Searle both appeal to the supernatural. Their downfall is ultimately the result of their insistence on distinguishing human intelligence. The Turing test measures intelligence, however, the capabilities of a specific form of intelligence are used to generalize the rest. The question of if AI can be intelligent can be rephrased as a question of whether AI can be more intelligent than humans in all forms of intelligence. Artificial Intelligence has already surpassed human intelligence in many aspects (chess, calculations, pattern recognition), and in other ways (language, music, art) it hasn’t. While we currently triumph in tasks that require more “creative intelligence,” machines are constantly improving. The potential of Artificial Intelligence is incomparably greater than human intelligence. Machines evolve through new hardware, updates, and technological advancements, while humans are restricted to the slow change of biological mutations. It is only a matter of time before machines have completely surpassed us in all aspects of capability.
Sources
- Costello, Katie. “Gartner Survey Shows 37 Percent of Organizations Have Implemented AI in Some Form.” Gartner. Gartner, January 21, 2019. https://www.gartner.com/en/newsroom/press-releases/2019-01-21-gartner-survey-shows-37-percent-of-organizations-have.
- Rimol, Meghan. “Gartner Forecasts Worldwide Artificial Intelligence Software Market to Reach $62 Billion in 2022.” Gartner. Gartner, November 22, 2021. https://www.gartner.com/en/newsroom/press-releases/2021-11-22-gartner-forecasts-worldwide-artificial-intelligence-software-market-to-reach-62-billion-in-2022#:~:text=Market%20Growth%20Will%20Accelerate%20as,new%20forecast%20from%20Gartner%2C%20Inc.
- Descartes René, and Donald A. Cress. Discourse on Method and Meditations on First Philosophy. Indianapolis: Hackett, 1998.
- Turing, Alan M. “Computing Machinery and Intelligence.” Mind 59, no. 236 (Oct. 1950): 433–450. http://www.jstor.com/stable/2251299
- Searle, John R. “Minds, Brains, and Programs.” Behavioral and Brain Sciences 3, no. 3 (1980): 417–24. doi:10.1017/S0140525X00005756.
- Boden, Margaret Ai: Its Nature and Future Chapter 6: But is it Intelligence, Really? pp 119–146; 28 pages
