As computers get “smarter” every year, I think it is interesting to look at the state of the art and summarize how we humans perform against AI, task by task. First of all, we are going to look at the main components of a computer’s hardware and compare them to our brain. Then, we will see how that translates into an ability to use real intelligence.

So let’s jump into it:

Energy Consumption

The brain is an expensive organ in terms of energy consumption: on average, your brain consumes about 20 percent of your body’s energy, and that’s in a resting state. The brain needs glucose to provide energy in order to process and transmit information through electrical signals, that’s why we perform worse in cognition demanding tasks if we are glucose depleted.

And by the way, it is a myth that you only use 10 percent of your brain. The bulk of your neurons are relatively silent for long stretches of time, waiting to spring into action when activated. But they’re doing so to remain energy efficient. If you somehow tripled the number of neurons activated at one time, the brain’s oxygen needs would increase dramatically, perhaps using as much as your leg muscles would while sprinting.

That being said, our brain needs 10 times less energy than the average computer. In fact, it requires about 10 watts which is even less power than a lightbulb, while a home computer runs on approximately 100 watts.

Storage

A typical computer will have just about 500GB of storage while gaming computers often require an SSD for additional storage, easily reaching 1TB or 2TB. The exact number of neurons in the human brain is not clear, but estimations range from 100 billion to as much as 200 billion, (see this Stanford University study).

You might be thinking, “Wait, the computer has bytes and the brain has neurons. How do we compare the two?”

Well, unlike computer flash memory, your neurons can combine with one another to assist with the creation and storage of memories. As I mentioned in previous articles, each neuron is connected on average to 1000 other neurons.

With over a trillion connections in an average human brain, this overlap effect creates an exponentially larger storage capacity, that can be estimated at around 1 petabyte. Even if this estimation is not exact, it is the equivalent of over a thousand 1TB SSDs, so human brains are still way ahead of computers in terms of storage capacity.

To put this in perspective, according to Computerworld, Yahoo — the Internet giant — has created a specially-built 2.0 petabyte “data warehouse”. Yahoo uses the immense information storage capacity of this data warehouse to analyze the behavior of its half-a-billion monthly visitors. “It is not only the world’s single-largest database, but also the busiest”, the magazine reported.

Although there are currently computers available on the market with more than 2TB if you are willing to pay the price, experts believe in the future, the maximum required for a single computer will stop around 16 TB.

Memory

So the human brain has significantly more storage than an average computer. And a computer can process information exponentially faster than a human brain.

How about accessing memory?

Well, here things get blurry. It depends on what kinds of information we’re talking about.

If a computer “knows” something, that information will always be accessible when needed. A human, on the other hand, may forget things unintentionally over time. However, computers lag behind humans in the ability to assign qualitative rankings to information. For a computer, all information is exactly the same. Humans, on the other hand, have many different types of memories and prioritize memories based on their importance.

Humans also relate memories to one another, computers lack this ability, at least for now.

Processing Speed

There is no contest here. When it comes to adding two plus two, computers are king. The fastest synaptic transmission takes about 1 millisecond.

Thus both in terms of spikes and synaptic transmission, your brain can perform at most about a thousand basic operations per second, or 10 million times slower than the computer.

Introducing Intelligence

Cool, so now we know that basically, what computers can do better than us is being fast. But is it the when it comes to real tasks, are humans still more intelligent than AI?

Well, before we can decide about that we should reflect a bit about what is intelligence. In my previous article, I argued it is a process that involves 3 stages, dominated by different skills: attention, creativity and logic.

By now, you probably have heard already that AI has already beaten humans at all these things: playing games like go, chess or poker, making accurate medical diagnoses, transcribing audio, reading lips. Even creative tasks like making music are already close to being mastered by machines.

But let’s think for a second when Deep Blue beat Kasparov at chess, what does it really mean? was that machine smarter than him?

Well, I think it is pretty clear the algorithm behind Deep Blue was written by some human or a group of them. So was the algorithm behind the poker bot Libratus, and so were all algorithms.

So these algorithms can know only as much as the people who wrote them.

Do software programmers know more than Kasparov about chess? Do they know more than doctors about diagnoses?

I think they obviously don’t.

The problem is the first stage of the intelligence process I described: attention. When a computer knows the same thing you know, you are going to make mistakes, you are going to get tired, you are going to lose focus. The computer won’t make any mistakes and will be always faster than you.

Present and the Future State of AI

I am a programmer and it is clear to me, that computers can beat us at any task that a programmer can teach it to do. I can fit a regression line by hand but a computer will do it faster and better for me. I can run, I can think of cool recipes for making cakes, I can tell a cat from a dog…you name it and if someone can write the code for it, the AI will do it better than me.

Robots can move pretty well already, all they need to do that is someone who knows physics to program them:

But still, we haven’t created yet a robot that can pass a Turing test, that is, it cannot pretend to be a human without us realizing it.

Because as I said, there are AI who can run, others can diagnose, others play chess and others can write. But humans run AND diagnose AND play AND write and do thousands and thousands of things.

That’s why AI has not reached human intelligence yet and is still far from doing it. They beat us at single tasks by speed and efficiency. We beat them at anything they don’t have data available to rely on. To make a human-like AI, programmers would need to add thousands of algorithms and then provide the software with a mega-hyper-huge database to store information.

A human who is say, 30 years old, has been learning things for 30 years. How much storage capacity do you need for an AI to have accessible in order to emulate 30 years of human experience? of tastes and noises, of pleasures and dislikes, of successes and failures.

Conclusion

Perhaps, the comparison of humans vs AI is after all useless. Instead of doing the massive job of writing bigger-than-life code for a robot to be human, we should use AI in an intelligent way, that is, the help us where we need it.

In my opinion, the future of AI in the coming years will not be to replace humans, not at all. It will be to fuse with us humans to enhance our abilities. Brain chips could provide humans with enhanced cognition, they could help power our brains with electric current where we need it. They could make our focus better, our thinking faster, and more efficient. That’s the direction I would like to see AI take in the coming years and decades.

Let’s not fear AI and embrace it as a way to improve our lives and expand the abilities we already have.