AI is advancing faster than ever, but is it truly reaching the level of human intelligence? Let us explore what AI can do, where it still falls short, and whether we have already crossed into a new era of advanced machine intelligence.

By Necip Gurler, PhD


The Journey from Turing’s Vision to Today
In 1950, Alan Turing posed a groundbreaking question: “Can machines think?” His seminal paper, Computing Machinery and Intelligence, laid the foundation for artificial intelligence and introduced the famous Turing Test, a benchmark designed to determine whether a machine could exhibit behavior indistinguishable from that of a human. Turing’s vision urged humanity to imagine machines capable not only of calculation but also of cognition—a bold idea that continues to inspire and challenge us today.
Nearly five decades later, in 1997, IBM’s Deep Blue defeated Garry Kasparov in chess, marking a defining milestone in AI history. Yet, as remarkable as this victory was, it showcased AI’s strength in solving narrow, structured problems, which remains distinct from the nuanced, adaptive, and creative nature of human intelligence.
Today, generative AI tools like ChatGPT, Claude, and Gemini are performing tasks that once seemed the exclusive domain of human creativity: writing essays, composing music, and creating art. But how close are we to achieving human-level intelligence in AI? The answer depends not only on how we define “intelligence” but also on how we envision our future alongside these increasingly capable machines.

Redefining Intelligence: A Dual Perspective
Defining intelligence is deceptively simple yet profoundly complex. Psychologists and AI researchers have proposed numerous definitions, but consensus remains elusive. A prominent synthesis by Legg and Hutter defines intelligence as “an agent’s ability to achieve goals in a wide range of environments.” This reflects two core aspects of intelligence: task-specific skill (“achieving goals”) and adaptability (“wide range of environments”).
In AI, this duality shows up in the contrast between specialized systems (like chess-playing engines) and those designed for broader learning across diverse challenges. This raises an essential question: Can machines ever replicate the interplay of these forms of intelligence that humans demonstrate daily?

What Does AI Excel At Today?
Modern AI systems demonstrate remarkable capabilities that often rival, and sometimes surpass, human performance in specific domains. For example:
◗ Specialized Problem Solving: AI systems like AlphaFold have cracked challenges like protein structure prediction, a major advance for drug discovery and medicine.
◗ Linguistic and Creative Outputs: Generative models can craft essays, engage in nuanced conversations, and even create art, marking significant progress in natural language understanding.
◗ Pattern Recognition at Scale: Deep learning algorithms have revolutionized how industries approach image recognition, fraud detection, and real-time translation, outperforming humans in speed and accuracy.

Notably, OpenAI’s experimental model o3 recently scored 87.5% on the ARC-AGI test, a benchmark measuring progress toward artificial general intelligence (AGI), a term often used interchangeably with “human-level intelligence”. This result demonstrates AI’s ability to generalize better than ever before. Similarly, advancements in reinforcement learning and multi-modal AI hint at a future where machines can seamlessly integrate visual, auditory, and textual inputs.

Where AI Falls Short
Despite these extraordinary advancements, AI still struggles to replicate the breadth and depth of human intelligence. Key limitations include:

  1. Common Sense and Intuition: AI systems lack the mental models animals, even a household cat, use to interact with their environment. Developing these “world models” is a crucial step toward true intelligence.
  2. Generalization Across Domains: Humans easily transfer knowledge from one area to another. AI systems, however, can stumble when given tasks that fall outside their training data.
  3. Memory and Learning: Unlike humans, AI lacks persistent memory and hierarchical learning. A four-year-old child grasps object permanence through simple observation—an ability that remains elusive for current AI.
  4. Emotional and Social Intelligence: While AI can detect sentiment, it cannot authentically interpret or respond to human emotions. Empathy and ethical decision-making remain uniquely human traits.
  5. Consciousness and Self-Awareness: The question of whether AI can ever be conscious remains one of philosophy as much as science. Current systems, no matter how advanced, operate without subjective experience or self-awareness.

Conclusion: The Road to Human-Level Intelligence – A Shared Responsibility
AI’s journey toward human-level intelligence is a testament to human ingenuity, but it also brings profound challenges. As AI systems increasingly match or surpass human-level performance in certain tasks, the responsibility to develop and deploy these technologies responsibly becomes paramount.
Governments and organizations must create regulations and guidelines that align AI progress with public safety and democratic values. Without shared responsibility, the risks of misuse or unintended consequences loom large.
So, are we at human-level intelligence yet? Most experts see no theoretical barriers, though the timeline remains uncertain, with predictions ranging from a few years to several decades. When AGI arrives, it may do so quietly, revealing its transformative potential only over time. The real question is whether we will have implemented the right frameworks to ensure that advanced AI benefits humanity as a whole.
I began with Alan Turing’s famous question, “Can machines think?” It is fitting to close with the insights of Turkish mathematician Distinguished Prof Cahit Arf, who asked a similar question in his 1959 seminar in Erzurum: “Can Machines Think and How?” Arf observed that while machines might one day match humans in speed and precision, they lack the ability to process aesthetic and uncertain elements—qualities crucial to human intelligence. He envisioned machines that could integrate atomic-scale uncertainties to emulate human creativity and judgment but doubted whether such breakthroughs would ever be fully realized.
Today, Arf’s vision intersects with the field of quantum computing. By leveraging principles like superposition and entanglement, quantum machines thrive on uncertainty, offering a potential bridge between human and machine cognition. Could quantum computing be the key to fulfilling Arf’s vision? In my next article, I will explore this groundbreaking field and its implications for the future of intelligence. ■

Necip Gurler is the VP of R&D at eKare Inc., where he leads the development of AI-based and multi-sensory medical imaging products for wound care. Previously, he held technical and executive roles at a defense company in Türkiye. Some of the projects he led include handheld landmine detection and LIDAR-based speed enforcement systems. He holds a PhD in electrical and electronics engineering from Bilkent University. He is currently an EUREKA Expert and serves as an Industrial Advisory Board member for the Electrical Engineering Technology program at ODU.