A friend of mine who works in AI was venting to me the other day about something totally absurd.
He said, “Ask AI to solve calculus problems, play Go, or write code? It crushes it in seconds. But ask it to sense the awkwardness in a conversation, soothe a crying stranger’s kid, or cook a meal in a kitchen it’s never seen before? It’s completely clueless.”
I said, “Well, isn’t that just common sense?”
He replied, “Common sense my foot! This is actually the most famous ‘Moravec’s Paradox’ in the field of Artificial Intelligence. Basically: tasks that humans find hard, AI finds easy. But tasks that humans find easy, AI finds harder than climbing a mountain.”
Back in 1988, scientists thought, ‘Hey, if we can get computers to do calculus and play chess, getting a robot to carry a tray or sweep the floor must be a piece of cake, right?’
But once they actually tried, they realized: writing a chess program takes a few thousand lines of code. But getting a robot to walk steadily across a messy room? That drained all the computing power of the top labs at the time, and they still couldn’t do it.
01 Moravec’s Paradox
My friend gave me a great analogy.
He said it’s like a rich kid (a ‘second-generation’ heir). Watching his dad and a team of old veterans run the business, he thinks taking over will be a breeze. But the moment he actually takes the helm, he realizes the things those old pros do effortlessly are impossible for him.
Think about it. Our sensorimotor systems—seeing, hearing, balancing, hand-eye coordination—are the ‘old guard’ that evolution has been polishing for millions, maybe even hundreds of millions of years. They are optimized to perfection.
Meanwhile, our logical reasoning and abstract thinking—arithmetic, chess, writing—are the ‘new department’ established only about 70,000 years ago. The processes are still pretty rough.
But for computers, it’s the exact opposite. Logic and calculation are their factory settings. To simulate the instincts humans spent millions of years evolving, they need massive amounts of computing power.
He asked me, “Have you noticed? We humans have this ‘intellectual arrogance.’ We think things you need school to learn are ‘high-level intelligence,’ and things we are born with are just ‘low-level instincts.'”
But Moravec found out: there is no such thing as a low-level instinct. That is god-tier code written by nature over hundreds of millions of years, using countless lives as beta testers.
The logical reasoning humans are so proud of? It’s just a small patch added recently in the history of evolution.
02 The Smarter AI Gets, The More Valuable You Become
Okay, so what are the things you do well that are a nightmare for AI?
McKinsey predicts that by 2030, 400 to 800 million jobs globally will be replaced by automation. But the ones getting hit the hardest aren’t blue-collar jobs. It’s the highly structured, predictable white-collar work—data entry, junior accounting, customer service, and even parts of legal and medical work.
Ironically, AI can’t touch plumbers, hairdressers, or nannies in the short term.
Computing power will depreciate; perceptive power will appreciate.
So, if you’re currently anxious about “Will AI replace me?”, ask yourself these three questions:
- Does your job require different operations for different scenarios? Take roof repair, for example. Every broken roof is different. You have to solve problems on-site, probably while dealing with an auntie who won’t stop nagging. AI can’t do that gig.
- Does your job require knowing the client to succeed? Think about major account sales. You have to read the hesitation in the other party’s eyes during a bid, or catch a subtle hint during negotiations. AI can’t give you that nuanced info.
- Does your job require continuous innovation and aesthetics? Like writing, design, and creativity. AI can mimic, and it can occasionally generate new stuff, but it needs you to curate and pick the good stuff.
If your job fits any of these three, congratulations. You are doing the hardest thing for AI to replace.
The smarter AI gets, the more valuable humans become.
But this “value” isn’t because you calculate better than AI. It’s because you are more human than AI.
So, what is the one thing you do that AI just can’t steal? Let’s hear it in the comments.