In the ever-evolving landscape of artificial intelligence (AI), public discussion grapple with a myriad of questions. Is AI truly intelligent? Can it be creative? Does it possess emotions? Awareness? Is it a dangerous black box? We seem to have endless discussions about these topics both at the dinner table and in the media. It adds to the mystery of AI, but it really isn’t warranted. The mystery, if you can call it that, is actually in our language.
The crux of the matter isn’t a lack of understanding about AI; rather, it’s a challenge rooted in the language we use to define and describe these concepts. Take intelligence, for example. When debating whether AI is intelligent, the conversation inevitably reflects the debate about the definition of intelligence itself. Much like pondering the intelligence of a dog or a horse, the discourse becomes a labyrinth of definitions and perspectives, showcasing the inherent complexity of the term.
Intelligence has been defined in many ways: the capacity for abstraction, logic, understanding, self-awareness, learning, emotional knowledge, reasoning, planning, creativity, critical thinking, and problem-solving. It can be described as the ability to perceive or infer information; and to retain it as knowledge to be applied to adaptive behaviors within an environment or context.
Wikipedia
Take above definition of the term intelligence from Wikipedia. Not only does it show the many competing ways intelligence has been defined, it also illustrates that the various words used to define it are themselves ambiguous. For example, what is “capacity for understanding” exactly? You can dig further into this term on Wikipedia and all you will find is more ambiguity. With these many ways to define intelligence, it’s impossible to ever settle a discussion about whether AI is intelligent or not.
The same intricacy applies to the term creativity. Because what exactly is creativity? Even consulting reputable dictionaries leaves us with explanations reliant on further intricate terms like “originality”, “novelty” and “creation”. We end up in an endless vicious circle of vague and ambiguous definitions. Attempting to draw a clear line between creativity and its variations becomes an exercise in navigating a linguistic maze that has no exit. The truth is that there is no definitive answer. There are no experts or authorities out there who can settle the matter.
The term emotion, too, is a battleground of definitions. The “does AI have emotions” debate isn’t fundamentally about how AI works or what it is capable of; rather, it revolves around the absence of a universal definition for emotions. Like with intelligence, these discussions resemble the never-ending debates we can have about whether certain animals have emotions. The same can be said of labels like “awareness” and “consciousness”.
The discussions go beyond terms that we usually use about living creatures. For example, the assertion that AI is a “black box,” shrouded in mystery, is another contentious statement. While the lack of visibility into the inner workings of AI models is often lamented, it can be argued that every calculation within a neural network is technically transparent. The computer performs each calculation, and theoretically, we can trace every step from start to finish. This perspective challenges the notion of a black box and introduces the idea that AI is, in fact, a “white box” – transparent but exceptionally complex.
In the grand scheme, much of the debate surrounding these questions tends to be circular, leading nowhere and echoing age-old discussions. The real challenge lies in the inherent complexity of the concepts themselves. AI is not a realm where easy answers reside; it is a space where nuance and understanding thrive.
Alas, the answer to whether AI is intelligent (or creative, or conscious, or…) is not mysterious at all. It’s in fact very simple: It depends on how you define the term. Once you decide on a clear unambiguous working definition of one of these complex terms, you can evaluate AI. Or rather, a specific implementation of AI – because some models may fall within your chosen definition while others may not. Not all AI models are equal.
The irony of it all is that we often want to use these labels to demystify and understand AI but end of transferring the ambiguity of our language to AI. In the process we make AI seem more mysterious than it really is. This is not to say that there are no aspects of AI that needs to be demystified: Even experts can’t fully comprehend why large neural networks behave exactly as they do. Not because of ambiguity of terms or a lack of transparency, but because of the sheer complexity of these immense models. But that’s a topic for another article.