Observing how people casually use the term “AI,” I decided to deep dive in AI terminology to explore the concepts of intelligence and whether we can label existing modern so-called AI solutions as true AI.
Many may think they know what AI is since they develop it themselves. However, I urge you to reconsider and look at it from the perspective of someone encountering AI for the first time, then compare it with your current understanding of AI; you will see a lot.
First, we need to grasp the scientific perspective of AI. AI is a branch of computer science with two objectives: 1) studying and modeling the intellectual processes of the brain, and 2) solving intellectual tasks more efficiently than the brain does.
Intellectual processes of the brain constitute a part of its informational processes. However, not all informational processes can be considered intellectual. For instance, simple memorization or algorithmic processes, which we do constantly—calculations in our minds or searching for information in certain sources—are not typically considered intellectual processes.
Now, what is intelligence? There is no single definition, but let’s take the abilities of intelligence according to Nickerson, Smith, and Perkins, the authors of the book “The teaching of thinking”:
- The ability to classify images: When we see two pictures of the same person, from a computer’s perspective, these images may appear different due to changes in perspective or facial expressions. However, we still say that it’s the same image, classifying it into the same category. This is one of the fundamental abilities of human intelligence.
- The ability to learn.
- The ability for deductive reasoning, meaning the skill to draw conclusions from existing facts.
- The ability for inductive reasoning, which involves generalizing based on specific facts. Both these types of thinking allow humans to go beyond current information and generate new knowledge, which may not always be correct.
- The ability to develop and use conceptual models, general representations of the world, one’s country, or one’s position in the family or at work.
- The ability to understand.
Let’s take another example, where intelligence encompasses the following set of abilities:
- The ability to organize knowledge based on relevance and establish relationships of relevance between existing data and knowledge.
- The ability to use logic in a broad sense, that is, to reason logically.
- The presence of reflection, meaning the capacity to evaluate one’s activities, draw conclusions about one’s behavior, and learn from it.
- The ability to learn new knowledge and correct previously acquired knowledge.
- The ability to generate hypotheses and possess cognitive curiosity, since incentives (often external-life enforces) compel us to gain knowledge, even if it may not be of use tomorrow.
Now, think about it. How many of the existing AI systems possess such capabilities? In my opinion, none. Yet, we casually call them AI. As suggested by one of AI experts, all current solutions might be more accurately labeled as Artificial Communication systems.
However, in general, it doesn’t matter how we call systems that mimic human intelligence. The question is what kind of AI do we need? IMHO, AI should be universal and transparent to help people to reveal their intellectual-creative potential. I will write more about this in the next article.
So…WDYT?