Are you smarter than a computer?
As of now are you smarter than a computer?
I have to be honest — I can’t give a definitive answer without it being useless. So, my goal is not to give you one. My goal is to give you a framework that helps you think better about human-made tools and their limitations. From slingshots to airplanes to computers, we will try to understand the scientific principles that govern them and their constraints defined by physics. Only then can we understand how smart a computer currently is and how much smarter it can become. If you agree to these rules — let’s go back in time to start our journey!
Since the “dawn of mankind,” humans have faced countless challenges. One of our earliest challenges was obtaining food. Initially, humans relied on muscle power, using stones to injure or immobilize animals. With the confidence bestowed by our frontal lobes, we sought to create tools to improve our lives. To better obtain food, we invented slingshots, utilizing the elasticity of leather to hurl stones more effectively. We did this by using available materials and our understanding of them. The more we understood, the more tools we could build.
Our ambitions and problems soon extended far beyond basic survival as we continued to improve our lives. We started using animals like horses as a means for transportation. We saw other animals that go faster. We called them cheetahs (cheaters). We yearned for greater speed. However, the complexity of biological cloning, or simply the domestication of such wild animals, was beyond our grasp at that time. What we understood were steam and steel. So, we built trains of steel that rolled on wheels instead of running on limbs, and later, cars that operated on combustion. These machines were not versatile enough to feed themselves or procreate to make baby cars, but they could easily surpass a cheetah in a drag race. When we wanted to fly like birds, we didn’t make machines that flap their wings. Instead, we employed Bernoulli’s principle to generate lift in airplanes. We solved the problems of survival and beyond way differently than the solutions that were bestowed on us by nature.
Every object in this world, whether made by man or nature, is governed by the physics principles they operate on. The laws of physics dictate their capabilities as well as their limitations. Supersonic planes can travel 10 times faster than the fastest birds — their shape is designed to be more aerodynamic, and the materials they are built on can withstand greater air resistance. Similarly, cars, designed for endurance and speed, can travel longer and faster than a cheetah. Unlike animals that suffer from muscle fatigue, machines don’t tire.
A computer is governed by a different set of physics principles compared to the human brain and possesses different limitations. While computers utilize silicon-based transistors and binary arithmetic for processing, humans rely on the brain’s complex neural networks and are comprised of biological tissues like cells. This difference enables computers to be quick at processing large amounts of data. Imagine you work at a store whose sadistic boss always gives you a list of all the items sold for the day and asks you to calculate the total revenue manually. You might be in a lot of pain, but a computer can effortlessly get it done in milliseconds (they Excel, don’t they). It wasn’t always like this, however.
Early computers used mechanical gears and levers to execute instructions, which made them slow. However, computers’ abilities have been improving with the advent of new materials and our understanding of new principles. Computers today use electricity to transfer data, making them lightning-fast. They are also more scalable: you can combine multiple computers to increase their capabilities, unlike human physical abilities, which have limited improvement potential. For instance, you can’t keep building muscle at the gym to run faster or increase your mental computation ability by eating almonds (south Asian people will understand). Computers and humans follow different improvement trajectories. While the limits of computer advancement are still unknown, humans in their current biological form will remain more or less the same. Unless Elon has his ways and we become Neuralink cyborgs.
If this makes you feel tiny, let’s look at tasks that humans are good at. Evolution has made us adept at pattern recognition, which was necessary for our survival (I can’t imagine storing the crop yields for 20 years was the top priority for survival). We perform this far better than computers, taking in multiple inputs from our senses to form intuitions and estimates. We can look in the distance and recognize our friend with his distinct walk or sound, and we immediately know it is Alex Earl of Stockton — and we should avoid him as a threat. Computers are still struggling to match this level of efficiency in human-like recognition. The human brain accomplishes all it does with just 20 watts of power. How can a circuit board with silicon transistors compare to a network of neurons, refined and optimized over millions of years of evolution?
领英推荐
What if we draw inspiration from nature and create a computer that mimics the human brain, one that simulates multiple neurons firing simultaneously to solve problems? Would such a computer match human intelligence? This idea is in fact not new. It was explored in the 1940s and implemented in the 1950s. However, it initially fell short due to the physical limitations of computers at that time. Even though computational tasks can be represented in abstract terms with mathematics, the implementation of that computation is a physical process. Every computation you make needs to be tied to something physical. Originally, gears were used, followed by vacuum tubes, and now transistors. Each of these possesses distinct computational capabilities, while all of them differs to the biological brain, which consist of neurons and functions through electrochemical processes.
The first successful attempt to simulate a human neural network was made by Warren McCulloch and Walter Pitts in 1943. Although this experiment did not come close to replicating the complexity and capabilities of the human brain, it marked a significant beginning. The computation carried out on silicon chips, was slow and fundamentally different from the brain’s chemical signals and multiple neuronal firings. After 50 years of progress, we now have GPT-4, which is way more effective at simulating certain aspects of human brain functions. Its success is attributed to Nvidia GPUs (Graphical Processing Units) which offer a computational approach more akin to the brain’s natural processes than traditional computing technologies, which could only execute one command at a time.
If you’ve witnessed what modern computers can do and are trying to compare them with your own intellectual abilities — might I be so bold as to say that, it is a pointless conversation? Let me explain why. A car outpaces a cheetah using different physical principles and manifests motion in a different way. Similarly, computers that simulate the human brain end up manifesting intelligence differently than their biological counterparts. Your chatbot has a higher bandwidth than us humans. It can read a book in a matter of seconds. In addition, we can connect multiple of these computers to make a bigger computer (neural net), a feat not possible with the human brain. However, human intelligence is more versatile — we efficiently process various types of sensory inputs (smell, touch, taste) and make judgments with minimal energy. To give a conclusion of all I have said so far, I believe that computers will indeed surpass human intelligence, but it is not going to be in a linear fashion. Comparing the human brain and a computer is like comparing apples and oranges. The comparison isn’t about a computer being 40% or even 50% as capable as the brain. It might perform certain tasks 1000 times better than a human brain and others 1000 times worse. Computers, like us, have their limitations. It’s at the edge of these limitations where humans can sometimes find their niche and, in fact, outperform our tools. At least for now. :)
“We like to believe that Man is in some subtle way superior to the rest of creation. It is best if he can be shown to be necessarily superior, for then there is no danger of him losing his commanding position…It is likely to be quite strong in intellectual people, since they value the power of thinking more highly than others, and are more inclined to base their belief in the superiority of Man on this power” — Alan Turing (1950) on “Can Machines think?”
The frontier of these limitations in the realm of intelligent machines is indeed narrowing, but the advancements are remarkable. Here’s an overview of the latest developments:
With computers, we are taking a different route to intelligence, so we need to be creative with what we understand in representing data, learning techniques, and doing computation on physical hardware. There is so much that is still yet to be done.
“We may hope that machines will eventually compete with men in all purely intellectual fields. But which are the best ones to start with?… We can only see a short distance ahead, but we can see plenty that needs to be done.” — Alan Turing (1950)
Great works that made this writing possible.
Want to give a shout-out to all my amazing friends who took the time to read through multiple drafts of my work and provided their valuable feedback.