Two questions fascinate me: (1) Does life exist in the universe beyond the confines of Earth? (2) Will we ever create artificial intelligence here on Earth? If you seek an answer in movies, books, and art, you get a resounding “YES!” to both questions. However, the scientific and theological issues surrounding these questions are far more difficult to sort out. After writing a book asking Is There Life Out There?, I would now like to turn to the second question. As an introduction, we need to define some terms.
When I first started learning about artificial intelligence (AI), the term usually referred to self-aware robots or computer programs (think R2-D2, C-3PO, the Master Control Program, WOPR/Joshua, D.A.R.Y.L., Skynet/Terminator, Number 5, and the list goes on . . .). The same term has a more expansive meaning today. As I read the description of various AI classes, my mind arranges these classes in a progression of increasingly sophisticated functionality.
Just a Computer Program
One could argue that much of the motivation behind AI research is to transfer human tasks to machines. Thus, on the first level of AI classification, computer programs provide the entrance to this endeavor. Humans have designed programs to add numbers, track budgets, calculate rocket trajectories, organize books, and perform countless other jobs. Every program operates by taking some data, doing some preprogrammed operations, and outputting a result. To get a different result from a computer program, you must either feed it different data or change the rules for executing the operations.
Artificial Narrow Intelligence
The next rung on the AI ladder is a program that takes a range of input, determines the most efficient path to a goal, and outputs the result. However, this artificial narrow intelligence (ANI) program has the ability to adapt and learn from previous situations. Consequently, an ANI encountering an identical set of input can opt for a different path based on accumulated knowledge from past “experiences.” Weak AI and ANI are often synonymous terms.
Arguably, the most significant example of an ANI was IBM’s Deep Blue, the computer that first bested a reigning chess champion in 1997. Fourteen years later, IBM’s Watson beat two of the most successful Jeopardy champions. ANIs perform a small set of tasks at a level that matches or exceeds that of humanity but have virtually no ability to function outside that capacity. These types of AI have been around for decades. Beyond the competitive realm, voice-assistant ANIs like Cortana, Alexa, Siri, and Google Assistant interpret voice commands and perform appropriate actions to carry out the commands.
Artificial General Intelligence
A multitude of ANIs exist today, and putting many of them together enables machines to accomplish rather complicated tasks like driving a car. However, this functionality still falls short of artificial general intelligence (AGI) or strong AI. This class would encompass almost all of the movie AI examples listed above. An AGI can do almost everything a human does and will accomplish some of those things in a superior fashion (but often at the detriment of others). The key that distinguishes an AGI from an ANI is that an AGI actually thinks and has a mind. At best, an ANI only acts like it thinks and has a mind. Consider arriving home from work or school and addressing your family with “Good evening. How’s everyone doing tonight?” An ANI might recognize “evening” and “tonight” and even your cheerful disposition, associate those cues with the need for dinner, and ask if you would like it to order pizza and soda. In contrast, an AGI would recognize that you are in a good mood and wonder what might have happened during your day to contribute to the mood. Thinking a little celebration might be in order, the AGI asks if you would like some pizza and soda for the evening.
The results might appear similar, but the ANI arrived at the activity by association and training whereas the AGI navigated a thoughtful, rational path that considered what’s best for you. The increased functionality of an AGI means it incorporates “lessons” learned from many different environments and applies them in novel ways to completely unrelated situations.
Artificial Super Intelligence (ASI)
For the last 50 years, the processing power of computers has grown exponentially. Informally known as Moore’s law, the number of transistors contained in integrated circuits has doubled roughly every two years. That number started at around 1,000 in the early 1970s and has grown to more than 10,000,000,000. Assuming it’s possible to create an AGI, and if the exponential growth of computing power continues, the abilities of the AI could also grow far beyond human capacity. Such a scenario leads to an artificial superintelligence, or ASI—the highest level of sophistication in my classification. People thinking about this scenario envision ASIs with almost godlike power. They offer an analogy of what a mouse thinks compared to what a human thinks. By all measures, the mouse has no comprehension of the simplest things a human ponders. Similarly, the incredible computational and intellectual power available to an ASI would provide access to unimaginable knowledge and abilities.
Interestingly, most people assessing what the advent of an ASI would bring think one of two outcomes would ensue. Either the ASI deems humanity a pest or irrelevant to its goals and it exterminates us. Or the ASI produces innovations that result in human immortality.
How Should We Proceed?
I see a parallel between AI and the multiverse. Both terms have a variety of meanings and the level of speculation grows as one moves up the multiverse and AI ladders. Two types of issues surround the discussion of AI. First, from a technological/scientific perspective we can ask, what can we actually accomplish and how might those developments benefit or harm humanity? Second, from a theological/philosophical perspective we can ask, how does the development of an AI interact with Christianity? Does it support the biblical view of humanity and our relationship to God or does it undermine it?
I’ll explore these questions in various levels of detail in future posts For now, I hope this introduction provides a foundation for understanding and reflection.