Are computers smarter than humans?

answered by Fleur Zeldenrust

Editors, Translator: Bryan Souza & Sjoerd Murris, Björn van der Haas

Short answer

No. Although computers can outperform humans in certain tasks, they fall short in many others. Humans are more versatile, efficient, capable of reasoning, and able to think creatively – all qualities beyond the reach of computers.

Longer answer

Computers outperform humans at specific tasks. In 1997, a computer (IBM’s Deep Blue) defeated Garry Kasparov, the world’s top chess player at the time. Since then, computers have become increasingly skilled at various games. The program AlphaGo, for instance, mastered numerous games by repeatedly playing against itself. Among these was the ancient Chinese game Go, which many believed was too complex for computers to master. However, in 2023, it was discovered that humans could still beat AlphaGo by making simple, illogical moves. Computers are also better than humans at performing precise calculations, particularly when large numbers of detailed computations are required – such as in simulations for weather forecasting.

Computers thrive when a task is clearly defined. Everyday life, however, is far more unpredictable and complex than games with fixed rules. This unpredictability is one major reason why it remains so difficult to create self-driving cars that can navigate busy city traffic effectively. Consider the challenges: pedestrians, children playing in the streets, cyclists, and more. In the more predictable environment of motorways, these self-driving cars perform much better!

Humans, however, are far more adaptable and capable of handling a wider variety of situations. We are better at managing new, unexpected scenarios and adjusting to unfamiliar environments. This adaptability is literal – the human brain constantly adapts, even modifying its physical structure when needed. Computers cannot do this. Moreover, while game-focused computers are excellent at their specialised tasks, they are incapable of anything else. In contrast, humans can cook, fold laundry (a surprisingly difficult task for robots), skateboard, and even juggle.

Humans are more efficient than computers. First of all, humans use less energy. If a human used as much energy per calculation as a computer, they’d need truckloads of peanut butter sandwiches daily! Secondly, humans require far fewer examples (data) to learn. While computers need to process millions of examples before they can recognise a cat in an image, humans can achieve this with significantly fewer examples. This disparity further highlights why computers consume so much energy.

An AI-generated illustration (Midjourney) with the prompt: ‘Computers versus humans. Fight.’

Creativity and reasoning: Chatbots like ChatGPT and other so-called ‘large language models’ can do something long thought to be uniquely human and impossible for computers to master: producing language. When you chat with such a program, it feels almost as if you’re conversing with another person. This is significant because the so-called Turing Test – devised by Alan Turing, a founding figure in computer science – assesses whether a computer can convincingly imitate an intelligent human. Passing this test is considered a key indicator of artificial intelligence. However, this assumption has faced criticism (imitating intelligence isn’t the same as being intelligent!). So, does the ability of these language models to mimic humans mean they are highly intelligent? Ultimately, no.

What these programs actually do is analyse vast amounts of text from the internet to recognise patterns and the structure of language. This allows them to predict likely responses based on a few words or sentences. However, these programs do not reason. Similarly, so-called ‘diffusion models’ can predict new ‘art’ based on patterns in previously seen artworks. Examples include ‘Sora’ for video, ‘DALL-E’ or ‘Midjourney’ for images, and ‘Suno’ for music. Yet, these programs don’t aim to evoke specific emotions or provoke new ways of thinking. A computer program has never experienced anything it tries to share on an emotional level. It creates something that resembles existing art but is slightly different. This new class of computer models may appear, in some ways, to be highly human-like, but what they excel at is recognising patterns in human behaviour and expressions.

Further reading