There is particular fear about the scope of Generative Artificial Intelligence systems, such as ChatGPT, Midjourney or DALL-E. But the reality is that this technology has many more uses and ramifications that we have yet to fully observe in action. Such is the case of Swift, an AI for drone piloting that has outperformed human experts.
A serious obsession has grown in recent months over what jobs these platforms could take outright. The matter has reached such a degree that even Open AI He has spoken about the matter, listing those jobs that will be most affected by the arrival of this new era.
This is how graphic designers, mathematicians, journalists, illustrators and a long etcetera must adapt to the use of these new tools. While other trades, such as the operation of agricultural machinery, carpenters and mechanics will continue to be active without a serious impact on their daily work.
But what no one saw coming is the emergence of an Artificial Intelligence capable of putting professional drone pilots out of work.
This is Swift: the Artificial Intelligence that swept a drone competition
The newspaper friends Guardian has published details about an AI project developed by a group of researchers from the University of Zurichwhere they created a system called swift capable of controlling drone flights.
After months of development, the engineering team decided to put Artificial Intelligence to the test and set up a drone race in June 2022, where the platform competed directly against professional pilots of this class of ships.
The results of the project and the impressive competition were not made public for a long time, but have finally been published in a recent issue of the magazine Nature together with audiovisual material that demonstrates the real degree of performance of the AI:
The assembled circuit with an approximate extension of 75 meters had a series of 7 doors or frames that the drones had to go through without crashing. Of the 25 races that were held between Swift and the human drivers, the AI won at 15 of the tracks.
The AI was consistently faster through tight corners and on the start, reacting on average 120 milliseconds sooner than the human drivers. In addition, he accelerated faster and reached higher speeds entering the first gate of the circuit.
It should be noted that the AI competed against world champion pilots in this field, such as Thomas Bitmatta, Marvin Schäpper and Alex Vanover.
As we can see in the video, the drone piloted by Swift showed better performance when taking turns and controlling acceleration by a wide margin.
So the result of this experiment represents an important milestone for this field.