The Miracle and Menace of AI Part 4
No Man Behind the Curtain
The book Beyond Death by Gary Habermas and J.P. Moreland contains a useful thought experiment to illustrate AI called the “Chinese Room.” In this example, imagine an English-speaking man being put into a windowless room with pieces of paper that have Chinese characters written on them. The man, who does not speak Chinese, is given a comprehensive manual on how to order the various squiggles and slashes in relation to other squiggles and slashes. After enough practice, a Chinese person can slip questions or comments into the room and get back responses that are indistinguishable from those of a native Chinese speaker. The Chinese speaker assumes that the person in the room understands Chinese, but in reality, there is no comprehension – just programming.
The lack of true understanding in AI boils down to the fact that there is no known way to explain how consciousness (i.e. a mind or a soul) could arise solely from matter. It turns out there is a robust philosophical basis for the existence of minds that are not identical to our physical brains. This viewpoint is the dominant position of Christian philosophers and is referred to as mind-body dualism.
To illustrate the mind/brain distinction, try a little experiment. Close your eyes and think of a “cat.” Maybe you see an image of a black cat with golden eyes under a crescent moon, or just your neighbor’s tabby that digs up your flower beds. Perhaps you feel a fuzzy warmth or hear the thrum of a purring feline.
Those images or sensations do not have a physical home. If we could somehow crack open your brain and read every neuron, we would nowhere find that same image of a cat. That picture exists in your mind. Your mind is what experiences the image of the cat.
To an AI program, a “cat” is nothing more than a sequence of letters or phonetic sounds that is then coded into numbers. There is no concept of “cat-ness” that the AI is “thinking” of when you enter that word, any more than your solar-powered calculator thinks of “groceries” or “rent” when you add up your budgets for the month. The AI program is simply performing a sophisticated statistical operation.
Words, Not Truth
In a similar vein, it is vital to understand that AI programs are not designed to provide truth. Likewise, AI is not capable of discerning value, justice, or morality. If intelligence consists of “knowing” many things, wisdom can be defined as well-applied knowledge. AI programs can search the internet much faster than you can, so in one sense, they are more intelligent. But we all know people who are incredibly smart but not very wise. Simply having access to many facts does not impart wisdom, which is always in short supply.
When you submit a question to an AI chatbot, you’re really getting a sophisticated internet search that summarizes sites for you without you having to click on each individual link. If I flooded the internet with claims that “scientists discover the moon is made of green cheese,” AI would happily report that Earth’s satellite is a giant ball of congealed dairy product. If you are wise enough to be skeptical of conventional internet sources, you will also be skeptical of AI-generated answers.
Because there is no mind behind AI, it isn’t “reasoning” or using the rules of logic when generating a response. AI programs are mathematical algorithms resembling a very large, albeit sophisticated, Plinko game. This lack of rationality is quite easy to demonstrate by having two different people ask the same AI program the same question. Very often, two different answers are provided due to differences in individual search histories or preferences.
Even the creators of ChatGPT recognize this limitation. When I recently logged into a corporate account for the first time, an introductory window popped up warning me not to use the AI for advice.
So if you’ve read this far, I would like to ask you a favor. Never start a sentence with “Grok says” or “ChatGPT says.” Those programs don’t “say” anything. They just regurgitate information grabbed from the internet based on mathematical formulae, with a bit of randomness thrown in to sound more lifelike. Instead, take a little extra time and read the primary sources they are drawing from. You may find that those sources aren’t reliable, or that they represent a non-Christian worldview.
Captive Creativity
Ironically, the exponential increase in AI-generated content is expected to cause problems for AI itself. AI programs are heavily dependent on internet-linked content, which is used as the raw material for its probability-based output. Human-generated original content will become an increasingly smaller percentage of the worldwide web. What will happen when a feedback loop is created, where AI-generated content is being used to generate even more AI content?
Some researchers believe that AI programs may suffer model collapse, like what happens when a copy is made of a copy, over and over again. Real human datasets include occasionally improbable items, since unlikely or rare things occur from time to time. AI, on the other hand, slants toward what is most common. After multiple feedback loops, AI programs can “forget” that the rare datapoints exist, or they just blur them to some unrealistic average value.
But the most likely outcome is that the internet will be filled with “blahness.” Human-provided prompts can infuse some creativity into the process, but if you’ve looked at much AI-generated content, it has certain telltale markers. While the technologies may grow more refined, AI algorithms still must draw on their training data. That means the output will tend toward the average depiction of whatever training material was used. And “average” has never been a word used to describe great art, philosophy, or science.
In Parts 1-4 of this series, I risked mild embarrassment by predicting the future of AI and its impact on the world. If I had to sum up my predictions in two words, they would be “not good.” In Part 5, I will propose a survival plan for living in the age of AI.
Brandon Aldinger is a chemist with a doctoral degree who works in an industrial research laboratory. He’s had lifelong interest in issues of science and faith, and he is passionate about training fellow Christians to think clearly about and stand firm on their beliefs within a hostile culture.