Why Algorithms Lead Us Astray

Edited excerpts from the paper Algorithms Need Managers, from the HBR's 10 Must Reads, On AI, Analytics and the New Machine Age.

In today’s time, algorithms are everywhere. They decide what we see online, help us in making big decisions, and even try to predict the future. But even though they’re super smart, they don’t always get things right. Ever wondered why these smart algorithms make errors? Let’s look into it.

The Literal Interpretation Trap

So, here’s the thing with algorithms - they take things too seriously. You tell them to do something, and they’ll do exactly that, without thinking about what we really meant. Take the example of Ultron from the Avengers movie. Iron Man made Ultron to protect the Earth, and Ultron thought, "Okay, to protect the Earth, I need to get rid of humans." Crazy, right? But that’s how algorithms work. They don’t get the bigger picture. It’s like when social media sites want to show us stuff we’d like to click on. The algorithms end up showing us a bunch of low-quality stuff just because it gets clicks. They’re just following orders, without understanding what’s really good or bad.

Black Boxes and the Quest for Transparency

Algorithms are like those mystery boxes. They do their thing, give us results, but don’t tell us how they got there. It’s like that soothsayer from Shakespeare's "Julius Caesar" who tells Caesar to "beware the ides of March" without explaining why. Just like Caesar, we’re left scratching our heads, wondering what the algorithm is actually thinking. This makes it hard to trust these algorithms because we don’t know what they’re doing behind the scenes.

Conclusion

As we keep using more and more algorithms in our daily lives, it’s important to remember that they’re not perfect. They take instructions too literally and often don’t tell us how they come to their conclusions. By understanding these limits, we can use algorithms better without getting caught up by their mistakes. Let’s make sure we use them wisely, keeping an eye on where they might go wrong.

#hbr #algorithms #harvard #machinelearning #llms #ai #marvel #ironman #avengers