Can machines surpass human intelligence in cognitive capabilities?

In 1997, chess grandmaster Garry Kasparov engaged in a six-game rematch with the supercomputer Deep Blue. After defeating the machine the previous year with three wins, one loss, and two draws, Kasparov faced a tougher challenge this time. Although he won the first game, Deep Blue ultimately triumphed with two victories, one loss, and three draws.

In 2011, IBM's Watson, powered by over 2,800 processor cores and vast data resources, competed against Ken Jennings and Brad Rutter on "Jeopardy!" The computer's ability to understand and respond to natural language demonstrated a significant advancement in artificial intelligence. Today, Watson's capabilities are being applied to solve complex challenges in healthcare and other industries.

Did these defeats indicate that computers had surpassed human intelligence? While it's undeniable that computers can execute calculations at astonishing speeds—such as the Sequoia supercomputer's ability to handle 16.32 quadrillion floating-point operations per second [source: Top 500]—how does this compare to the gray matter in our heads?

Quantifying the speed of human thought is challenging, leaving us to rely on educated guesses. Robotics expert Hans Moravec from Carnegie Mellon suggests that humans process around 100 trillion instructions per second (teraflops) based on visual processing [source: Moravec]. Meanwhile, Chris Westbury, an associate professor at the University of Alberta, estimates the brain's capacity at 20 million billion calculations per second (20 petaflops), considering neuron count and signal transmission speed [source: Westbury]. While computer speeds are nearing or even exceeding human thought, the question remains: Are computers truly smarter?

As of now, computers lack true intelligence. But will this remain the case in the future?

Computers and the Human Brain

Intelligence extends far beyond raw processing speed. Although supercomputers like Sequoia can solve problems faster than humans, they lack the ability to adapt and learn as we do. Human brains excel at navigating unfamiliar scenarios, leveraging past experiences, and experimenting with solutions. Computers, on the other hand, require explicit instructions and cannot independently innovate or adapt.

Humans possess a remarkable talent for identifying patterns. While advancements in machine learning have enabled basic pattern recognition—such as facial recognition in digital cameras—computers still struggle with complex, nuanced patterns that humans can effortlessly interpret and adapt to.

Is it possible for computer scientists to develop a machine that mimics human thought? The challenge lies in the brain's extraordinary complexity. Without a complete understanding of how the brain works, creating an accurate simulation remains a daunting task.

For computers to surpass human intelligence, they would need the ability to derive conclusions from observations. In a 2009 study, Cornell University engineers developed a program that achieved this on a small scale. The software observed a pendulum's movements and deduced fundamental physics principles, accomplishing in a day what took humans millennia to understand [source: Steele].

Although the Cornell project marked a significant milestone in computer engineering, achieving computers capable of drawing conclusions from general observations remains a distant goal. The software provided the computer with the necessary tools for analysis, but it couldn't independently create or refine these tools.

As long as computers depend on pre-programmed instructions to execute tasks, they cannot be considered more intelligent than humans. Even advanced systems like IBM's Watson are limited to responding to inputs and cannot think or retrieve information spontaneously. True intelligence in computers will only emerge when they can adapt and perform tasks beyond their initial programming. Until then, they remain highly advanced calculators.

Numerous computer scientists are tackling this pivotal challenge. Some aim to develop computers that replicate human thought—a daunting task given our incomplete understanding of cognition. Others focus on creating systems that don't emulate the brain. Futurists such as Dr. Ray Kurzweil believe it's only a matter of time before we create self-aware computers capable of recursive self-improvement, enabling them to analyze and enhance their own performance.

However, developing a self-aware computer system is currently beyond our reach and may not even be feasible. As we advance in biology and computer science, we might face fundamental barriers to creating such machines. Alternatively, the distinction between human and machine intelligence could blur, rendering the question irrelevant.