Why AI will be Smarter than Humans — 100%

Published in

Artificial Intelligence in Plain English

6 min readMay 23, 2024

I believe there is essentially a 100% chance that AI will become much smarter than even the smartest humans, in pretty much any way you’d want to measure it.

(I’m not the only one who believes this, but I want to share a fresh perspective as to why I believe this is inevitable)

Some people believe that AI will begin to plateau, and will never surpass the smartest humans in areas where humans currently shine.

Note: Even Bill Gates has said that there are plenty of reasons to believe GPT technology has reached a plateau.

Will AI be smarter than the smartest humans?

Before we answer this, let’s have a little fun.

Forget about AI for just a second.

Let’s think about some examples where machines in general are so much better than humans it is actually hilarious.

Vision:

Imagine the person with the best eyesight in the world.

Compare that eyesight to either an electron microscope or a telescope.

Physical strength:

Imagine the strongest person in the world doing a deadlift.

Now compare that strength to a forklift.

Chess:

AI can crush even the best players in the world, every time. Experts used to believe this would be impossible.

This is a cartoon image of IBM’s Deep Blue, beating Garry Kasparov (world chess champion at the time) in 1997. This was a first, but certainly not the last.

In 2015, Google’s DeepMind also beat the best human at the game of Go, which is much more difficult for a machine to learn than chess.

Memory:

Imagine the person with the best memory in the world.

Compare that to a supercomputer.

LOL!

Calculation:

Who could calculate the square root of 5.9482434 to 8 decimal places faster — a math nerd or a supercomputer?

Again — LOL!

Speed:

Do you think we could make a machine move faster than a 100 meter gold medalist sprinter?

LOL!

Knowledge:

Who knows more facts — a professor or a supercomputer?

You get the idea.

Once a machine gets better than a human at something, there is almost no upper limit to how much better it can be.

Humans clearly have upper limits.

Our brains are only a certain size. Our eyes have limits. Our physical bodies have limits. etc.

If you like this, you’ll love our AI Growth Guys Newsletter

You will also like our AI YouTube Channel

But what about all of the things where humans are better than machines at?

Currently, there are many things where humans are much better than machines.

Humans are infinitely better than machines at tasks we find almost trivial.

Imagine this task:

Go into a kitchen and make a tuna sandwich, without making a mess.

You’d need to get out a can of tuna, open it, mix in some mayo. Get out some vegetables, wash them, chop them up…. Make some toast, put it on a plate, clean everything up…

Humans are much better at this task.

There are also many cognitive tasks where humans are currently much better than AI too.

Writing a great novel:

Again, the best novel writers are so much better and more creative than whatever ChatGPT could spit out, it isn’t even funny.

Creating the best music.

Although, AI is starting to catch up.

The best computer programmers can still beat AI easily.

Coding with ChatGPT is great, but there is a reason why the best programmers are still making millions per year.

Ok. I’ve given a few examples where AI or machines are far superior to the best humans, and vice versa.

We can easy imagine many more examples to put on each list.

So why am I so confident that AI will be able to outperform humans on every task, sometime in the near future?

Note: Maybe it will be 5 years, maybe 20 years. The exact timing doesn’t matter, the point is that the time is coming.

1. Multiple exponential curves coming together.

Compute

Nvidia is worth over 2 trillion dollars and has been pouring billions into making new and better chips. The increased investment in chip development and data centers is increasing exponentially. This will help AI get much better at everything.

Algorithm improvements

The models themselves, like GPT-4o are getting better and smarter at an exponential pace.

Data

The amount of data and the quality of data being fed into these models is increasing exponentially.

Note: Some say that we’ve already got all the low hanging fruit from the internet, but don’t forget about synthetic data, which is increasing exponentially. Also, we’re taking in data in different forms — read point #2 next.

2. Omni channel training will be huge!

Think about a simple task for a human. Grabbing a can of soda off of a table, opening it, and taking a sip.

It has been so difficult for machines because it’s virtually impossible to “hard code” the process into them. How can you describe exactly which muscles to use?

It is incredibly complicated, yet it is so easy for us.

But this will become much easier for AI, when we continue to feed a ton of data into them. The AI will be watching trillions of hours of video and learning what to do.

This is similar to what Tesla does with self driving. They don’t tell the cars what to do. They give them a bunch of sensors and a bunch of driving data, and say “you figure it out!”

In a lot of ways, this is how humans learn too. Think about catching a baseball. It is hard at first as a little kid (basically impossible), but it becomes easy (relatively) just by spending a few hours practicing with a sibling, friend, or parent.

The same thing with riding a bike. We don’t learn how to ride a bike based on a clear set of instructions of what to do. We need to practice, and fall down, a few times before we can do it ourselves.

We will be continuously feed LLM’s (Large language models like ChatGPT) all kinds of data — visual, audio, etc.

All of this data, combined with the power of the LLMs and series of exponential curves I mentioned above, will allow them to learn everything humans do.

Yes, it is true that this “last mile” might be more difficult than I think. For example, Tesla’s self driving is improving all the time, but it still hasn’t solved for those, once in a billion, edge case situations.

It will though, this is all still relatively new. Just wait a few years!

Think of how human-like AI voices are now.

ChatGPT’s new GPT-4o voice is very close to a real human voice. Not just the sound or the pronunciation either, but the emotion, the intonation, the understanding of context.

ChatGPT shocked the world in November 2022 with how good it was at writing like a human. It wasn’t perfect, but nothing before it was even close.
Sora, OpenAI’s generative video model is shockingly good (and getting better at a rapid pace)
OpenAI’s voice capabilities are shockingly good and human like.

It turns out that giving enormous amounts of training data and enormous amounts of compute can generate amazingly human like results. Much better than almost anyone (even the skeptics) would have predicted even a few years ago.

What is fundamentally human where no amount of training data or model improvements could be enough?

I think nothing.

I don’t feel bad about it either. We’re already comfortable with machines that are better than us in other ways, like a microscope for vision.

I’d love to hear what you think about this in the comments.

Check out our new YouTube Channel

Set up an AI consulting call with me: Andrew Best Minnect

In Plain English 🚀

Thank you for being a part of the In Plain English community! Before you go:

Be sure to clap and follow the writer ️👏️️
Follow us: X | LinkedIn | YouTube | Discord | Newsletter
Visit our other platforms: Stackademic | CoFeed | Venture | Cubed
Tired of blogging platforms that force you to deal with algorithmic content? Try Differ
More content at PlainEnglish.io