The human brain can pass synaptic impulses at a speed of roughly 120 meters per second. But the speed at which our brain can form thoughts is remarkably slow: just 10 bits per second, according to a new study.
Let’s put that into perspective: the average download speed in the US is around 200 Megabits per second: 200,000,000 faster. But it gets even weirder.
Your eyes alone absorb about one billion bits per second, processing vast amounts of visual data. Your ears, skin, and nose also receive an overwhelming influx of sensory input. Yet, at the level of conscious thought and action, humans operate at a glacial 10 bits per second. So, what gives? Why is our thinking process in slow motion?
Sensory Overload vs. Cognitive Bottleneck
Your nervous system is a master of parallel processing. All our senses are wired to handle massive and parallel data streams. But our brain is not designed to process all that data.
When it comes to conscious thought, decision-making, and action, everything becomes much slower. Even at peak performance, whether you’re typing, playing video games, or solving puzzles, you never seem to exceed 10 to 50 bits per second.
To look at why that is, Jieyu Zheng and Markus Meister, researchers from the Division of Biology and Biological Engineering CalTech, reviewed a century’s worth of experimental data. They looked at human reaction time tests, typing speeds, e-sports performances, and memory competitions. Using information theory, they quantified how much information is processed in various cognitive tasks, showing that this bottleneck is present across different types of human activity. From this, they calculated 10 bits per second.
“This is an extremely low number,” Meister said in a press statement. “Every moment, we are extracting just 10 bits from the trillion that our senses are taking in and using those 10 to perceive the world around us and make decisions. This raises a paradox: What is the brain doing to filter all of this information?”
Their findings suggest that human cognition operates at an incredibly slow rate due to limitations in neural architecture, evolutionary constraints, and serial processing mechanisms. While our senses process information in parallel (handling multiple streams at once), our conscious mind seems to operate in a strictly linear fashion — one thought at a time. That’s why it’s hard or impossible to truly be good at multitasking. For instance, try listening to two conversations at once. Even if they’re slow, it’s almost impossible. Three or four conversations are just impossible to tune in to. Your brain forces you to process ideas one after another, never simultaneously. This also explains multitasking: you’re not doing two things at once, you’re just switching from one to the other really fast.
Evolutionary Constraints
As frustrating as this limitation can be, it makes evolutionary sense. The brain’s primary job is not to think — it’s to move and keep you safe. Early lifeforms evolved brains to navigate their environment, find food, and avoid danger. As more advanced life forms (like humans) emerged these directives did not change; they just adapted. These tasks required simple, serial decision-making: move left or right, hunt or hide, fight or flee. In other words, your brain still operates on simplistic approaches.
Even as human intelligence evolved, our brains retained this one-decision-at-a-time model. We process complex thoughts the same way our ancient ancestors decided whether to chase prey — one slow step at a time.
This makes a lot of sense practically. If your brain processed all sensory information at full speed, you’d be overwhelmed. You’d notice every sound, every flicker of light, every background detail — all at once. Instead, your brain is extremely good at filtering out unnecessary information. It filters 99.99999% of incoming data and leaving only the few bits necessary for action. It’s not perfect but it protects you from sensory overload. However, it also bottlenecks your ability to think quickly.
AI Thinks Much Faster
The researchers highlight another important consequence of this finding: AI thinks much faster than us. It doesn’t have all the multi-layered information, it doesn’t have our heuristics and ability to select what’s important, but in terms of sheer, raw thinking power, it’s much faster.
AI models process information at kilobits to megabits per second — millions of times faster than human cognition. This is why AI can now beat humans at games like chess and Go. It isn’t because AI is “smarter” in the human sense — it’s simply thinking much, much faster.
The study dives into some rather depressing potential outcomes. In a world where AI can process millions of bits per second, humans may struggle to keep up. AI assistants will handle much of our decision-making simply because they can think faster. At some point, we may even become passengers in a world run by AI.
Think, for instance, about the driving infrastructure around us. All of it was built for our cognitive capacity and ability. But if AIs become drivers, then we can start changing infrastructure for the AI brain, not for the human one.
The discussion of whether autonomous cars will achieve human-level performance in traffic already seems quaint. Roads, bridges, and intersections are all designed for creatures that process at 10 bits/s. When the last human driver finally retires, we can update the infrastructure for machines with cognition at kilobits/s. By that point, humans will be advised to stay out of those ecological niches, just as snails should avoid the highways.”
Are the Days of Human Dominance Numbered?
I couldn’t help it and asked ChatGPT what it “thinks”. It said that AIs are indeed much faster at processing information than humans, but told me I shouldn’t worry because humans still have a place to play due to their creativity and ability to form long-term plans. But it also said the days of human dominance are numbered and that humans will be in charge in an “increasingly symbolic” way.
Granted, the AI is referring to humans as “us” which suggests it doesn’t have comprehension of what it’s saying. ChatGPT’s output is presumably not relevant in terms of comprehension. But it raises an important (and disturbing) question: are we truly headed for a future where human thought becomes secondary, falling behind what machines think? Are we headed for our own intellectual demise, conceding decision-making to artificial thought? Or is there still a key role we have to play?
Whatever the answer is, we’ll have to do it at 10 bits per second.
