A new mission to find Amelia Earharts plane turns to AI for help

For almost 90 years, the mystery of what happened to Amelia Earhart has endured. She and her plane disappeared somewhere over the central Pacific Ocean on July 2, 1937. Since then, it seems like every few years, an expedition sets out to find her plane.

The boats and recovery teams have always been armed with the latest equipment, historical diaries, and maps. I’m old enough to remember the 1989 TIGHAR expedition, the first modern civilian search to focus heavily on the “castaway hypothesis.”

Yes, some people thought she might still be alive, but most thought it was a big waste of time. That ’89 castaway trek, like all the searches before and since, returned empty-handed. Around eight years ago, there was a glimmer of hope that bones discovered in the Pacific were Earhart’s, but that study has been heavily disputed.

Well, here we are again. There’s renewed hope of finding Amelia and determining where she actually perished. A fresh team of researchers from Purdue University has arrived on the remote island of Nikumaroro.

The crew was featured on Wednesday’s Good Morning America. The segment noted that the island where they’ve set up their HQ is home to flesh-eating crabs whose claw power rivals the bite of a lion. This is a trip I would decline, even in the almost-impossible instance that I was invited.

Crabs aside, they are there to investigate a mysterious underwater shape in the island’s lagoon known as the “Taraia Object.” Once again, some are on pins and needles over whether the Earhart conundrum will finally be resolved so she can rest in peace.

There’s a twist this time, because AI was deployed. So, obviously, if the plane is found, those AI agents who banded together to hack Hugging Face will most certainly come together again and try to take credit for the discovery. But I digress.

In all seriousness, generative AI models like ChatGPT were prompted to analyze historical theories and place a digital pin where Earhart’s Lockheed Electra 10-E might be resting. Ironically, the algorithm pointed to the same visual anomaly that human scientists had spotted on satellite imagery years earlier.

At first glance, this all looks impressive, but ChatGPT didn’t analyze the ocean floor, for example. Instead, AI synthesized decades of human speculation. And there’s been a helluva lot of speculation over the last nearly 90 years.

Yet while the Earhart-AI connection makes for some flashy pop-culture and scientific headlines, it also represents something happening beneath the surface — literally, beneath the world’s oceans. Artificial intelligence is stepping out of science-fiction and becoming the ultimate tool for modern treasure hunters and maritime archaeologists.

We’re currently swimming in a golden age of deep-sea exploration, but, like everything else, it’s drowning in too much data and missing things, as The Conversation noted in March. For example, the ocean floor is estimated to hold up to 3 million undiscovered shipwrecks and downed aircraft.

So, you’re saying to yourself right now, “Well, it’s impossible that 3 million things can’t be found in the ocean.” But what if I told you that well over 70 percent of the Earth’s ocean floor remains unmapped in high-resolution detail?

What does this mean? I live in NYC, so I asked AI for a metaphor to visualize it. It explained that if New York City sat in one of those unmapped regions, the resolution would be so low that you’d see only a single, blurry shape where Manhattan used to be, while every skyscraper, neighborhood, street, and subway entrance would remain invisible.

Now you understand why 3 million ships and aircraft remain undiscovered.

When modern underwater drones dive into the deep ocean, they return with mountains of sonar data and millions of hours of murky video. For decades, human researchers had to painstakingly review every second of footage by hand. That’s a brutally boring process that took months, sometimes even years, and is ripe for errors.

That’s where AI comes in and says something like, “Years?”

Instead of waiting for a chatbot to guess a location, institutions like NOAA and the University of Michigan are using specialized AI built for one job: spotting anomalies directly on the ocean floor. Scientists feed these models thousands of hours of footage so the AI learns what the normal ocean floor looks like; reefs, sand ripples, cliffs, and more.

Once it knows the natural layout, AI can spot the telltale right angles of man-made metal and rotting wood. To someone like me, who has had a lifelong fascination with the ocean, this is incredible.

And so are the results! The U.S. Navy and NOAA recently tested a predictive model across the coastlines of the U.S. and Puerto Rico. The AI scanned the data and correctly flagged known shipwreck sites with 92 percent accuracy, showing how quickly it can shrink massive search areas.

What the Earhart experiment really shows is that intelligent machines are learning to see through the darkness of the deep, helping make the abyss more transparent. Amelia Earhart may have been lost to the limitations of 1937 technology, but if she’s ever found, it could be because tomorrow’s technology finally caught up with her.

This article represents the opinions of its writer.

John Casey is a columnist at Raw Story, the Advocate, Daily Beast, and Outsports. He’s also the author of Love: The Heroic Stories of Marriage Equality.