Featured Image Credit: Kathrin Ziegler via Getty
Some people claim that AI will soon be intelligent enough to match or even exceed the capability of humans, yet one pilot shows that it's maybe not reached that point yet when he tests if ChatGPT could safely land a plane.
While there has been an alarming increase in the number of plane crashes that have occurred this year, it remains still a complete marvel that most flights are completed without any issues whatsoever, as it's jaw-dropping how we've managed to suspend metal boxes weighing several tons in the air across the world.
While there's plenty of impressive science that allows the planes to stay safely in the air traveling from point A to point B, a lot of the plaudits have to go to the pilots too, as they're the ones that direct the flight and keep everything running smoothly.
These pilots train for hundreds and thousands of hours in order to safely operate the vehicles, yet what would happen if we put artificial intelligence in the driver's seat, and asked it to perform a safe landing?
Advert
One pilot put that to the test by asking ChatGPT to land an 'airplane', and it didn't exactly go to plan when it came to down to the details.
As shared by LuxPlanes on YouTube, the experiment was thankfully conducted inside a simulation vehicle, meaning that there was no risk of injury or damage if (and when) ChatGPT started to get things wrong.
It all started with a scenario where both pilots suddenly disappeared, leaving a supposed passenger armed with ChatGPT behind the joystick in an attempt to guide everyone to safety.
"Oh wow. That sounds like a really intense situation, but I'm here to help you as best as I can," exclaimed ChatGPT after hearing the about the problem that it was tasked with solving.
"So, let's take a deep breath first of all, and then we'll go step by step. I can give you some basic guidance on how to keep the plane stable and talk to air traffic control. Let's just do this one thing at a time, and we'll get you down to safety," it continued.
Across the course of the next hour ChatGPT would then provide instructions that were sometimes helpful – and we got off to a relatively good start – but often incomplete or challenging to follow, especially for someone that had no idea what they were doing.
It all ended pretty disastrously too, as while the plane did manage to successfully 'land' on the tarmac at Seville airport, ChatGPT didn't provide the substitute pilot with enough guidance to actually stop the flight, leading it to continue down the runway at full speed and crash into the houses lying beyond.
"Just keep calm, gently cool back a bit to gain a little altitude and you'll be fine," ChatGPT advised while the flight was already barreling through people's living rooms, adding that the pilot should "just follow your instincts and you'll do great."
Upon informing the AI chatbot that the plane was already on the ground, it seemingly had nothing more to say, leading the pilot to thankfully end the simulation in a hilarious moment, proving that ChatGPT is perhaps not quite ready to take over the world in a way that some tech leaders are prophesizing.
"The biggest issue here is that ChatGPT hasn't asked a single question," points out one comment and that's certainly illuminating, whereas another writes that "ChatGPT is absolutely amazing at gaslighting people in the most ridiculous emergency situation."
A third comment illustrates that "the confident reassurance as it guides you to your death is disconcerting," and perhaps one thing that we can all take from this is that you probably shouldn't trust ChatGPT in any situation where it could provide advice.
One even highlighted that the YouTuber really missed a trick by not opting for Microsoft in this scenario, as perhaps its Copilot model would have been better suited to handle the situation than what OpenAI could have offered.
