I read about a flight that was required to land in the Hudson River in New York. US Airways Flight 1549 was an Airbus A320-214 which, in the climb out after takeoff from New York City’s LaGuardia Airport on January 15, 2009, struck a flock of Canadian geese, just northeast of the George Washington Bridge and consequently, lost all engine power. Unable to reach an airport, pilots Chesley Sullenberger and Jeffrey Skiles glided the plane to a ditching in the Hudson River off Midtown Manhattan. All 155 people aboard were rescued by nearby boats and there were few serious injuries.
This is the animation video of the incident. While reading about the incident and after watching the video and the video clip from the movie Sully (2016), questions came to my mind how far can a machine (based on AI- Artificial Intelligence-) take over human jobs and are they good enough to interpret the tricky situations. In this case, the incident occurred at the height of 2800 feet which is considered very low to handle such incidents. The pilots showed fantastic skills and managed to land in the Hudson River in very chilly conditions.
When various parameters were analyzed by different agencies, they used data from the black box and created simulations based on the data retrieved. Initially, they concluded that there may have been a pilot error because the auto simulations proved that the pilots should have acted differently and could have landed on either of the two airports which were in the vicinity. Both these airports had cleared the flight for emergency landing. But the pilot declared that he could not make it to the airports. Then data was loaded on simulators and two pilots “flew” the aeroplane, based on data captured. About 13 such pairs “flew” the plane on the simulator. Seven pairs could not reach the airport.
But when the pilot Sully was being interviewed, he said that he did not agree with the findings because the findings were applied to the situation, immediately instructed the pilot to turn towards the airport. But when the actual bird hits happened, the pilots were stunned and by rough estimate did not react for 35 seconds. The plane ultimately landed in the river 218 seconds after the bird hit. Out of this, 218 seconds 35 seconds were lost in the pilots recovering from the shock. Another important aspect was that the procedure stipulated by the plane manufacturer Airbus had written it with the assumption that the flying height would be 33000 feet at such times. At this height, humans have enough time to react. But at 2800 feet height, time left to react was less than four minutes before either landing or crashing. When 35 seconds of “no action” time were considered in the simulation tests, the results matched with the pilots’ actions in deciding that not enough time was left for them to go back to any of the airports. Both pilots were honoured by many authorities for their heroic efforts which saved all 155 lives. See this interesting clip below, from the Movie “Sully” based on the incident with Tom Hank in the lead road.
Now the question comes to mind who is better Man or Machine? Would machine have required 35 seconds of reaction time? Maybe no. But unless this real-life situation was available in the database for the AI system, how would a machine have reacted? The pilot could see that there was the river Hudson around, which he thought would have become a spot for “soft landing”. At any other place in New York or New Jersey, the aeroplane would have simply crashed into buildings and maybe exploded. Would an AI system have known at the time of the incident, the option for a soft landing? Another question that comes to mind is how many combinations would be required to be present in AI database, for the system to understand that combination of 2800 feet height, time remaining of (218-35=) 183 seconds, wide enough river Hudson being around and there were no taller buildings in the path etc, etc. By analyzing hundreds of such options, the AI system would have arrived at a conclusion, what would be the “safest” option to save lives and the aircraft. At least in this case, with the current level of technology, the human decision was a superior decision.
The way humans react, their mind thinks out of the box! AI system decides only based on what data is available with it. The human mind applies its thoughts and extrapolates to arrive at the correct conclusion. It automatically does the risk analysis and tries to take the best possible decision under the circumstances. How far AI systems will be able to extrapolate, it is difficult for us to judge. 50 years hence, who knows?
I will share another example from the Aviation field. The incident is known as “Gimli Glider”, this happened in 1983 and the plane was Boeing 767 used by Air Canada. There were series of errors in the measurement of fuel, manual, gauges, dip stick, at every stage. There was a transition going on from the FPS system to MKS system in Canada at that time. One thing led to another ending in the incident. See the link below.
Incident is too technical to be explained here. But the end result was all the fuel tanks were suddenly empty, half way to the destination. We can say that all these errors and complications probably could have been avoided by AI. Yes and no both because AI stuff takes time to “learn everything”! After many such incidents, AI would become “an expert” in this area. But what happened later is where human intervention will show us the limitations of AI over human intelligence. When the pilots felt some issue, they decided to do an emergency landing at Winnipeg. But when they realized that there was zero fuel, the copilot and the pilot started thinking fast. The pilot was a gliding expert and he calculated that the plane could glide 16 times the height at which the plane was flying. The copilot was from that area, and he had worked on a Royal Canadian Air Force Station in the area. He judged the parameters and decided to land at that Ex-air force station at Gimli. Only problem was that both the pilots and air traffic controllers were not aware that the station had become a racing track. But it was only because of the knowledge of the pilots and the way they applied it sensibly, the plane ultimately landed safely with no major injuries to anyone on the flight as well as on ground.
This brings out the same question, AI or Humans who are better? This discussion is going to go on forever but AI systems will become better and better but nobody really knows what is stored in the Human brain and what the humans will retrieve from their brain, is going to be really difficult to judge. But we should not forget one thing, humans are the ones who are creating an AI system!
I read somewhere about the movie Sully! The copilot Skiles is asked in the end what you would have done differently if the same incident were to occur! He smiles and says, “I would make sure that we landed in Hudson River in July and not as we did in January!”