In search of the essence of artificial intelligence
In the previous article we traced together the history of artificial intelligence "AI" (artificial intelligence), from the first neuron models to the modern era of generative AI. Today, I'd like to address a topic that has preoccupied researchers, philosophers and technology specialists for decades - how to properly define artificial intelligence?
This question may seem simple, but the answer is not obvious. Despite the fact that the term "artificial intelligence" appeared as early as the 1950s, to this day there is still debate over its precise meaning. Let's take a closer look at the issue and why formulating a unified definition of AI remains a significant challenge.
Why is the definition of AI a real challenge?
Artificial intelligence is not just a trendy buzzword, but a field that is constantly evolving and surprising. Although the term "AI" has become ubiquitous, there is no single, universally accepted definition. It's like trying to photograph the horizon - the closer we get, the further it moves! Why is this happening? First of all, AI is trying to mimic human intelligence, yet we ourselves do not fully understand what it is! Psychologists, neuroscientists and behaviorists have been studying the human mind for years, but they still haven't reached full agreement on the definition of intelligence. Since we don't fully know what human intelligence is, how can we accurately define its artificial counterpart?
Secondly, we often define AI by saying that it mimics "human abilities," but we don't specify what exactly we mean by those abilities. Is it about problem-solving? Creativity? Or perhaps the ability to learn? Definitions such as "technology that enables machines to mimic diverse and complex human abilities" or "the ability of a computer to perform complex tasks in difficult situations" are too vague and do not clarify what specific abilities are meant.
Different definitions, different perspectives
In the world of AI, we will encounter many attempts to define the field, each with its own approach:
AI as an activity aimed at making machines intelligent
Nils J. Nilsson, a specialist at Stanford University, defines AI as "activity aimed at making machines intelligent, and intelligence is the characteristic that enables an individual to function correctly and predictably in its environment." This approach emphasizes the goal of making machines capable of acting intelligently.
AI as systems that exhibit intelligent behavior
The European Commission describes AI as "systems that exhibit intelligent behavior, analyze their environment and take actions - with a certain degree of autonomy - to achieve certain goals". This definition emphasizes the ability of machines to analyze and act autonomously.
AI as an imitation of human capabilities
A common definition that refers to AI as technology that mimics the way humans think. This approach is very popular, but, as we mentioned, it raises questions about what we actually mean by "human capabilities."
AI as "prediction machines"
Some go a step further and suggest that we should talk about "prediction machines" because the modern technology behind AI is all about predictive ability.
Frequently appearing terms such as "a certain degree of autonomy" in definitions are very vague. After all, the classic autopilot in an airplane also "perceives" (through its sensors), "acts" (correcting the course) and "learns" (adapting to the flight conditions), yet we wouldn't call it a manifestation of artificial intelligence!"
Context and "AI effect" - how our perceptions are changing
The way we understand AI is changing as technology evolves. What we once thought was an amazing manifestation of machine intelligence is becoming ordinary to us over time. We call this phenomenon the "AI effect." When a computer learns to solve a difficult task, we think, "aha, it's just computation," and stop seeing it as "real" intelligence. Nick Bostrom aptly pointed out that AI is anything that surprises us at any given time, and when we stop being impressed, we just call it software! Who would have expected that a facial recognition app, which was once a breakthrough, is now something we have in our pocket, in every smartphone?
Is AI a new kind of "entity"?
Some philosophers, like Daniel Dennett, go even further and argue that we should not compare AI to humans. AI is something completely new, they are oracle-like "entities" - they make predictions, but they don't have human characteristics such as personality, conscience or emotions. AI does what we do, but it does it in a completely different way. As Edsger Dijkstra put it, "Do submarines swim in the same sense as fish?" This question provokes deeper thought. Although submarines move through the water similarly to fish, their mode of operation and mechanics are fundamentally different. Fish use complex biological mechanisms evolved over millions of years, while submarines are engineering structures that operate on the principles of physics and mechanics.
Similar is the question: "Do airplanes fly like birds?" At first glance, the answer seems simple - after all, both float in the air. However, when you take a closer look, you'll see that birds use complex biological adaptations - wings, feathers and muscles that allow them to fly with precision, while airplanes use engines and aerodynamic shapes designed by engineers.
These Dijkstra analogies perfectly illustrate the problem with comparing AI to human intelligence - despite similar results, the path to achieve them can be quite different.
Which definition is therefore best?
In the previous article I adopted a simpler definition of AI as "systems that can learn, adapt and make decisions, mimicking human thought processes to some extent." However, after an in-depth analysis of various perspectives and the latest positions of institutions such as the IEEE and ACM, I see the need to reformulate it. I now believe that the most general and understandable definition of AI is: "systems that exhibit intelligent behavior, analyze their environment and take actions - with a certain degree of autonomy - to achieve specific goals." This definition not only covers all current applications of AI, but also leaves room for its future development.
Remember that in the world of artificial intelligence we distinguish, among others, generative AI (GenAI), which creates new content, and general AI (AGI - Artificial General Intelligence), which strives to achieve a level of intelligence comparable to that of humans in a wide range of fields. In this article, we mainly focus on the definition of AI as such, which relates more to AGI.
Reflections on the path to understanding AI
Defining artificial intelligence is a fascinating, though not easy, adventure. The lack of a one-size-fits-all definition is not an obstacle, but a challenge that forces us to constantly reflect and analyze. Context, technological advances and philosophical considerations all affect how we understand AI. Let this article be an invitation for you to continue discussing and exploring the ins and outs of artificial intelligence. I hope you found this journey as inspiring as I did!
Thank you for being with me on this amazing adventure in the world of AI! Don't forget to share your thoughts in the comments!
See you in the next article!
Note!
This article has been prepared with the support of Claude 3.5 Sonnet, an advanced AI language model that has helped organize and present content while maintaining an authentic message and educational value.This article has also been automatically translated from Polish using DeepL. If you find any errors, please let me know in the comments.
