A Beginner’s Guide to AI: Machine Superiority

There is currently no machine capable of superhuman intelligence feats. I’m sorry to be the one bursting your bubble, but for now, that’s how the universe works.

It may seem like AI is smarter than humans in some ways. For example, the powerful neural networks used by big tech can search millions of files in seconds, a feat that would cost people more than a single life.

But that is not a superhuman intellectual achievement. It trades human attention for machine speed. On a file-by-file basis, there is no attention-based task that a machine can beat a human being at the exclusion of chance.

Greetings humanoids

Sign up now for a weekly recap of our favorite AI stories

Let’s take computer vision as an example. You can train an AI model to recognize images of cats. If you have 10 million images and you need to know roughly how many of them contain cats, AI should be able to do in seconds what would take humans many years.

However, if you took 100 school-aged children and had them look at 100 images to determine which images had cats, no neural network in existence today could consistently outperform the children.

the consideration

In this particular task, we can point out that: AI struggles to identify topics when the subject is not posed in a typical pattern. So, for example, if you had a cat dressed up as a clown, the average kid would still understand it’s a cat, but an AI not trained on cats dressed as other things could struggle or fail at the task.

The trade-off we’re talking about, when you’re sacrificing accuracy for speed, is at the heart of the AI ​​bias problem.

Imagine if there were a statistically significant number of images in our 10 million deck that depicted cats dressed as clowns and the model we used wrestled with costumed felines.

The cats-dressed-as-clowns may be in the minority of cat images, but they are still cats.

This means that unless the model is particularly robust against clowning, the entire population of cats dressed as clowns will likely be omitted from the AI’s count.

Of course, it would be trivial to train an AI to recognize cats dressed as clowns. But then there’s another fringe case where you have to train the AI, and another after that.

A human child does not have the same problem. You can dress a cat however you want, and as long as you have not blocked the identifying features of the cat, the child will understand that it is a cat in a costume.

This problem, on a large scale, is exactly why driverless cars are still unavailable outside of well-regulated testing.

Machine superiority

Machine superiority can be described as a paradigm in which computers know what is best for humans and have the ability to control us through nonviolent means.

When legions of killer robots suddenly travel back in time to subdue us with laser guns, that’s a different kind of superiority — one beyond the scope of this nonfiction article.

The big idea behind machine superiority is that one day we will create an AI so intelligent that it surpasses humans in its ability to reason, plan and adapt.

The reason it is such an attractive concept is twofold:

  1. It imagines a future where humans have finally solved all the world’s problems… by inventing a magic box to solve them for us.
  2. We can tell scary “what if” stories about what would happen if the magic box was bad.

But at the moment there is no machine that can outperform the average human at any given task, except for digital speed and chance.

When humans and machines are given a task that involves hidden information — for example, playing chess or determining whether someone is a Democrat or Republican by looking at a picture of their face — the ability to predict the next segment in a linear output can often prevail. have intelligence.

Instruments of the trade

Here’s what I mean: it doesn’t take intelligence for DeepMind to beat humans by chance or Go, it requires strict adherence to probability and the ability to predict the next move. A human can look at a chessboard and imagine X number of moves in a given time. An AI can theoretically imagine all possible movements in a given time.

That doesn’t mean the AI ​​isn’t one of the most incredible feats of computer science humanity has ever delivered, it just means it doesn’t have a superhuman level of intelligence — it’s nowhere near human level, in fact.

And there is not a human being in the world who can look at another’s face and determine their politics. We can guess, and AI can often guess more accurately than we can. But people’s faces don’t magically change; their politics often does.

A good way to look at artificial intelligence is to see it as a tool. It is easier for a human to cut down a tree with an ax than without one, but it is impossible for an ax to do anything without a human. The ax has superhuman resilience and tree-cutting abilities, but you wouldn’t call it superhuman.

An AI that can spot cats very quickly in images is a good tool if you ever need to spot a cat, but it’s not something you’d call superhuman. The same goes for someone who can beat us in games.

Machines are great at certain tasks. But we are still a long way from the emergence of a supreme digital being.

We’ll have to figure it out for ourselves, at least for now. Fortunately, we have AI to help us.

Neural’s “A Beginner’s Guide to AI” series started way back in 2018, making it one of the longest-running series on the fundamentals of artificial intelligence out there! You can view the rest below: