What AI Can't Do (Yet) 🤖

Artificial Intelligence is No Match for Natural Stupidity

Ah yes, more content on Artificial Intelligence. If you’ve been following the news, it seems like AI can do anything. They can put together business expenses, edit videos for you and try to talk you out of your marriage (Check out this NY Times Article if you want to know more). As the whole world has been flipped upside down in the last 6 months ever since ChatGPT was released, everyone has suddenly become an expert in large language models, neural networks and OpenAI. This isn’t a post about how to leverage the latest AI tool or how we’re all going to lose jobs to ChatGPT. Here, we’re talking about what makes you unique and different from Artificial Intelligence.

It’s Go Time

If you’re like me, the name Lee Sedol means nothing to you. But in 2016, Lee went up against AlphaGo (Deepmind’s Go playing AI). It was on the level of Garry Kasparov playing against DeepBlue. Lee Sedol, who has 18 world titles to his name, lost the set, 4-1 resigning in all of his losses. This stunned the world as Go is often regarded as the game that computers couldn’t solve.

With its breadth of 250 possible moves each turn (go is played on a 19 by 19 board compared to the much smaller eight by eight chess field) and a typical game depth of 150 moves, there are about 250150, or 10360 possible moves.

Christof Koch - Scientific American

This progress was further compounded when Lee Sedol retired in 2019, citing that it didn’t matter if he was the best in the world because there would be an AI who was always better than him.

So what now? Just ~3 months ago, Kellin Pelrine, who is technically one level below amateur rating (chump), beat the best AI Go player 14 out of 15 times. How in the world did he do it? Well, with some help from his own AI that analyzed KataGo’s weaknesses, he put some pieces on the edge of the board that distracted KataGo from his main plan. He added that “As a human it would be quite easy to spot” but because it was not a tactic that KataGo had previously seen, it was not easily accounted for. This is the beauty of being human. Let me explain.

Why Don’t We Have Self-Driving Cars?

In David Epstein’s Range, he discusses the difference between kind and wicked learning environments. Kind environments, are activities such as Chess or Golf where you are likely to face continual patterns with clear defined rules and boundaries that provide immediate feedback. The problem? We live in a world where most activities are wicked, not kind.

What does this have to do with AI? Computers, and more specifically AI systems thrive in kind environments. When they are provided with clear instructions within set parameters, they can produce absolutely spectacular results. We’ve seen this at play with ChatGPT. With a large database of knowledge and access to the internet, writing a poem based off noodles in the style of Robert Frost is child’s play. But put them in a wicked learning environment and it can cause a lot of problems.

Self-Driving cars have been a great example of this, they’ve been “just a year away” for the last 10 years. Why has it been such a problem? Because driving (especially in cities such as San Francisco), is a wicked environment. There are too many factors and variables that come into play when we drive. Pedestrians, construction, and bikers are too unpredictable for an artificial intelligence system to keep up with. That has nothing to do with other drivers who are already frustrating enough to deal with.

Side note: There are many funny videos that have come out of the confusion of self driving cars. These videos are comedic but have grave and serious consequences when you realize these are affecting lives.

Humans, on the other hand, thrive in wicked environments. Our ability to learn and apply across a wide range of spectrums is uncanny. Epstein calls this “range”

Our greatest strength is the exact opposite of narrow specialization. It is the ability to integrate broadly.

David Epstein - Range

Our range is able to learn in multiple different environments and transfer knowledge between disciplines is unique. There are random ideas that pop into our heads that we love to think about, learn about and want to explore. This strength is what separates us from computers and artificial intelligence. We’re a species that has an insatiable desire to learn and grow. Don’t believe me? Check this out.

Impossible Google Numbers

If you’ve ever doubted the genius of Larry Page and Sergei Brin, here’s a stat that should change your mind. There are over 8.5 billion Google searches a day. That’s over 3 trillion searches a year! But here comes the crazier stat.

And if you don’t believe me, Google it

Despite the billions of searches that Google gets every single day, there are roughly a billion queries that have never been seen before. What does this mean? Humans have a capacity for creativity that we cannot begin to understand. The overwhelming amount of new and creative ideas that we Google is just a small glimpse into our world. Sure, an AI can create poems, write code and generate a never before seen image, but those all come from our imagination.

We’ve already seen that AI struggles with any wicked learning environment. Creativity is the essence in thriving in wicked environments. As humans, we specialize and flourish in wicked environments because we’re able to generalize. We can take our past experiences that may be completely unrelated and apply them to our current situation. AI may be able to make a new Drake song, write a sonnet about waffles or create photos of entirely new animals. But AI can’t (at least for now) be creative. What can you do to get ahead? Get creative.

Derek

If you liked this piece and want to stay up to date with future posts, hit that button below!