Dr. Olgun Sadık

This is a translation of a blog post from 2022 with corrections and added details. Lately (2022), I’ve been reading and learning about artificial intelligence and its applications in education. I plan to share what I’ve found—without relying on academic language or worrying about formal in-text citations.

Let me start with a disclaimer: I’m not an expert in artificial intelligence. Beyond a Prolog course I took during my undergraduate studies (I mostly remember it being a language used specifically for AI), my background in this area is limited. Still, with some experience in computer science and programming beyond that of a general reader, I feel comfortable commenting a bit. Of course, I could try to sound authoritative with fancy statements, but the truth is: I’m still learning.

That said, working at Indiana University’s Department of Intelligent Systems Engineering, I can’t help but notice a major trend—many decisions, particularly those based on data, will increasingly be made by computers on our behalf in the future. Through what I’ve read and experienced, I hope to share some insights that others may also find interesting. At the very least, I believe blog posts get read far more often than formal academic articles.

I was inspired to begin this journey by my daughter—she’s currently into comics—so I started with a highly illustrated book titled Artificial Intelligence: A Graphic Guide. I should say upfront: while the visuals aren’t particularly impressive, the book presents AI concepts with brief and understandable definitions. It seems that interest in AI has been steadily growing over the past 50 years. The computer-vs-human chess matches many of us watched on TV were among the first public showcases of AI attempting to challenge the human brain. Even so, it’s clear that such challenges remained very limited as recently as 2022.

Driven by curiosity, I went beyond the book and did some research. I discovered that chess matches between humans and machines go back further than the 1990s, which is when most of us became aware of them. I learned that the idea that machines could possess some form of intelligence—enough to compete with humans—originates with Alan Turing. He was the mathematician who first proposed that machines could perform complex and time-consuming calculations that humans would struggle to complete, essentially laying the foundations for modern computing. His 1950 paper Computing Machinery and Intelligence addresses many of the questions that still confuse us about machine intelligence.

Though many chess matches between computers and humans have taken place over the years, the first major competition was in 1996 and IBM’s Deep Blue defeated Garry Kasparov in the second game in 1997. Reportedly, Deep Blue could calculate 200 million different positions in a second. Thinking about that processing power—even by 1990s standards—is pretty exciting. But at the same time, given that the number of possible decisions in chess is limited (even if the limit is vast), it’s debatable whether this actually qualifies as true intelligence.

As I jumped from article to article, I came to realize that while the developments in AI are exciting, we are still in the very early stages. Despite many promising forecasts for the future, today’s AI systems have major limitations. For instance, self-driving vehicles may misinterpret fake road signs.

This illustrates a key limitation: while today’s AI systems are capable of impressive feats, they tend to make decisions within the boundaries of the information they’ve been trained on. When current AI systems encounter unfamiliar obstacles—scenarios that fall outside their training data—they often struggle to respond appropriately. This is particularly true for systems that rely heavily on supervised learning, where knowledge is tightly coupled to labeled examples. That said, there are exciting developments in areas like reinforcement learning and meta-learning, which aim to improve adaptability and enable learning in more open-ended environments. Still, even these approaches have limitations: agents may overfit to simulated environments or require vast amounts of trial-and-error to generalize effectively. So while adaptation is an active area of AI research, achieving robust, real-time flexibility remains a major challenge.

This also shows that while machine learning—especially in its current forms—has led to remarkable progress in pattern recognition and task-specific optimization, it may still be a limited path toward general or adaptive intelligence. Today’s systems often excel at narrowly defined problems (e.g., image classification, chess), but they struggle with broader cognitive abilities like abstraction, analogy, or transferring knowledge to unfamiliar domains. In that sense, machine learning is powerful but doesn’t yet capture the kind of flexible, context-aware reasoning that we typically associate with human or biological intelligence.

By contrast, even the simplest living organisms are equipped with instinctive behaviors encoded in their genes—behaviors that allow them to adapt to new situations without prior learning. These include reflexes, innate drives, or sensorimotor patterns shaped by evolutionary pressures. For example, sea turtles instinctively crawl toward the ocean minutes after hatching. This kind of “built-in” intelligence gives biological systems a crucial survival advantage in new or unpredictable conditions. It raises a compelling question: Can AI systems ever develop something similar—an ability to respond meaningfully to unknown environments without having seen them before?

All this suggests that while machine learning is a powerful tool, it alone may not be sufficient for developing true general intelligence. We still need to understand how to enable broader reasoning, transfer, and flexible adaptation—capacities that biological systems display even in the absence of training data.

What excites me most is this question: where do living beings store the information that helps them adapt to new situations, and how do they transfer it across generations?

Think about it—during the COVID-19 lockdowns, how did people across Turkiye instinctively rush to stock up on dry goods? When were we taught that, and how has that instinct carried through to today? If these questions don’t spark some curiosity in you, perhaps you haven’t spent enough time thinking about them yet.

I realize my ideas here aren’t presented in a highly structured way. That’s because I’m just beginning to explore this topic and understand its connections. But if I stick with it, I’m hopeful that I’ll be able to share more exciting insights in a more organized way over time.

Reference: https://olgunsadik.netlify.app/blog/yapay-zeka-nedir-günün-şartlarında/