A human and AI discussion on helicopter safety (2024)

This article is dedicated to my friend R.C.

This article is about the confusion between Human and AI authorship in linkedin. I kindly ask the reader to join me in a voyage of discovery, bewilderment and awe as I did it, step by step.

My first concern was the fact that the question is based on a statement which is, at least, debatable: “helicopters have a higher crash rate as compared to planes”. “Where are the statistics? Please substantiate” were my initial issues while reading the post.

These are the answers provided by Marc Dawodu [MD henceforth in the text] – whose curriculum vitae and profile as provided in linkedin suggest a “Global Safety & Aviation Speaker, Marketing & Management Expert” as stated – that puzzled me:

“Here are a few reasons. Helicopters have a higher crash rate as compared to planes for several reasons, such as the following:

* Helicopters fly at lower altitudes and speeds than planes, which exposes them to more hazards and obstacles, such as power lines, buildings, trees, and birds.

* Helicopters also have less margin for error and recovery in case of an emergency, as they cannot glide or divert to a nearby airport like planes can.

* Helicopters have more complex and delicate machinery than planes, which requires more maintenance and inspection. Helicopters have a main rotor, a tail rotor, a transmission, and an engine, all of which are essential for flight and control. Any failure or damage to these components can cause a catastrophic crash.

* Helicopters are harder to fly and control than planes, especially in hovering and low-speed situations. Helicopters require constant and precise input from the pilot to maintain stability and direction, as well as coordination of the cyclic, collective, pedals, and throttle.

* Helicopters are also more susceptible to aerodynamic phenomena, such as vortex ring state, retreating blade stall, and ground resonance, which can cause loss of lift or control.

* Helicopters are used for more risky and demanding operations than planes, such as search and rescue, medical evacuation, law enforcement, firefighting, and military missions. These operations often involve flying in adverse weather, low visibility, high altitude, or hostile environments, which increase the chances of accidents.”

I do agree with some of the statements, other seem hollow to me and, above all, there is no statistical substantiation for such a claim. If one compares helicopters with fixed wing General Aviation, which means comparing aircraft with similar Maximum Take-Off Mass [MTOM] the difference in crash rates is attenuated (see references later). It’s the preponderance of airliners and Air Transportation in aviation that leads to a crash rate statistics that looks unfavourable to helicopters, either by measuring total hull losses or by measuring accidents per mile flown, and I could not find this in the answer.

The next surprise I had was finding that the post had 345 comments by then. I acknowledge I didn’t’ read all of them and a substantial fraction of the ones I read reflects fears or prejudice of their authors, other are humorous and should not be considered for learning purposes, but one question from Ozan E. again drew my attention:

“Such an interesting article and thanks for sharing.

I am wondering about ‘Could drones replace helicopters’. Do you have opinion about it.

Thanks in advanced,”

The post by MD was illustrated with a crash of a helicopter and some of the readers commenting said that all people aboard had died, others that no one had died; other asked which accident was depicted in the photo. I did a quick check with aviation-safety.net to find the answer at once: CFIT Accident Eurocopter EC 135P2 N558MT. This reminded me that linkedin is a community of people with a mosaic of professional backgrounds and that many people who comment are not necessarily educated in the field of expertise they are writing about.

Reading through the accident description, I learned that the main cause for Controlled Flight Into Terrain [=CFIT] was a human error, which seemed an odd and biased choice to support the initial MD’s statement-inside-a-question, since most aviation accidents involve at least some concurrence of human errors.

My initial reaction was to do nothing but by the end of the day I returned to the page and it had already 405 comments! By comparison my articles have 0 to 1 comment on average. Being such a popular discussion, I decided to contribute and answered the following to the person who asked about the accident:

To Ozan E., I replied:

“I think UAS (aka "drones") will replace Helicopters partially, depending on the concept of operation. It makes a big difference if you’re considering remotely operated UAS or autonomous UAS but it is too complex to put that in a reply.

I’ll leave you three examples with increasing helicopter value:

1. Supply delivery in hazardous environments. There is no point in putting a human into jeopardy if only materials need to be carried. Thus, send an UAS.

2. Search and Rescue: If you need to cover a wide area than you could use a swarm of UAS coordinated with a smaller number (or single) of helicopters: once the victim is found, the helicopter crew makes the rescue. Thus, send both.

3. Medical evacuation: if one or more passengers on board need medical assistance during transport, then it is very likely that the competences of a human pilot on board to overcome all inflight ambiguities and unknowns will offset the risk involved in adding an expert into the helicopter crew. Thus, send a crewed helicopter.”

I kept on reading comments and the most extraordinary finding was that at least four persons suggested that the initial post by MD was generated by an AI bot: “The ChatGPT is strong with this low effort post!” and the other three I read were similar. It is true that the text has the expositive form of Generative AI outputs, but this is also the form of a professor summarizing lessons.

I let it down for two weeks and returned to it on January 3rd, 2024 when the number of comments had raised to 690 and reposts exceeded 200. Then, I and a colleague made trials asking Bing (via Slype) and ChatGPT: “Why do helicopters have a higher crash rate as compared to planes?”. This is when the real topic of this article really began.

To summarise results, the initial Marc Dawodu [MD] contains six arguments. Bing’s response is in Annex 1 and contains five arguments.

#1 argument of MD coincides with #1 argument of Bing.

#3 argument of MD is somewhat coincident with #4 argument of Bing.

#4 argument of MD is the same as #2 argument of Bing but MD includes more information.

#6 argument of MD coincides with #5 argument of Bing.

The best added value of Bing is in the final explanation [see Annex 1] which includes figures, statistics and links for those who want to substantiate or extend their knowledge on the topic.

ChatGPT’s response is in Annex 2 and contains eight arguments.

#1 argument of MD coincides with #2 argument of ChatGPT.

#3 argument of MD coincides with #5 argument of ChatGPT while both add more information in different matters. MD describes critical components while ChatGPT mentions the importance of maintenance (which is a key part of continued airworthiness).

#4 argument of MD coincides with #1 argument of ChatGPT.

#6 argument of MD coincides with #6 argument of ChatGPT.

The best added value of ChatGPT reply is in the short texts that explain each of the arguments. I’d like to emphasize that ChatGPT transforms the initial statement by MD into an hypothesis and argument #8 refers the small size of the helicopter sample compared to fixed wing aviation. Both concerns look positive to me. Last but not the least, the messages that safety is in constant improvement and causes of accidents are usually multiple are also very noteworthy.

Given the high level of overlap between MD text and Bing and ChatGPT responses I put forward the hypotheses that the post was created by Bing, ChatGPT or something similar. To test such a hypothesis, I queried three on-line services that analyse the origin of the text and they return the following:

• copyleaks.com indicates “AI Content detected” for the whole MD text.
• zerogpt.com concludes 100% AI GPT generated, indicating the full 1549 characters and 235 words as an issue of AI/GPT.
• zerogpt.me concludes that “This text is likely to be a mix of human and AI” but there is a probability of 87% that it is fully written by AI and its 12 sentences are likely AI generated.

I used my own reply to Ozan E. as a control text to submit to the same services and their findings are:

• copyleaks.com indicates “This is human text”.
• zerogpt.com indicates 0% AI GPT generated.
• zerogpt.me concludes that “This text is most likely to be written by a human” and that there is 0% probability this text was entirely written by AI.

These results confounded me deeply but as I was realizing its broader significance, a lot of questions rushed in to my head and prompted me to write this article:

1. What if linkedin (owned by MicroSoft, which is also a reference shareholder of OpenAI, the creator of ChatGPT) modeled a Marc Dawodu out of “digital clay”, breathe life into it and sent him asking questions to boost dialogues on the platform?
2. Is there a human Marc Dawodu that asked ChatGPT or other similar AI tool that question, parsed the replies to boost visits to its profile?
3. Are there people or organizations “planting” their beliefs in the audiences with sentences like “There are only two types of helicopters: those that already crashed and those that will crash” to foster certain businesses or to hinder other businesses?
4. Is the hypothetical artificial origin of the post that makes it a magnet for comments and reposts?
5. What if commentators are also bots or people paid to “personify” multiple users?
6. How many “non-human” profiles are there in linkedin? What’s their purpose? Are there organizations that keep “phantom” profiles “alive” to satisfy people willing to pay? Cui bono?
7. What if linkedin is doing it internally to compensate the relative disaffection of the human users towards the contents in the platform?

Conclusion

By way of conclusion and going beyond the Marc Dawodu post, the following questions came to my mind:

1. Is it worth to continue investing time and money into linkedin?
2. Is it worth sharing expert knowledge to create a more informed community without knowing who is human beyond the people we know physically?
3. Is it worth to think in writing in articles such as this one?
4. Is it worth to participate in the collaborative articles or are we educating AI robots with our professional resources?
5. Is it worth to keep posting, replying to unknown users without knowing if is there any human left to learn from the discussion?
6. How can one find out the truth about this?
7. Is it worth to publish a linkedin article asking these questions?

Final remarks

The main topic of this article is the Human and AI authorship of contents in linkedin. Nevertheless, the trigger for it was a discussion about helicopter safety, a topic I sincerely care fare more about. Any reader interested in airworthiness and helicopter safety may read the following remarks. I’m done with the Human & AI topic until I find more material in the comments below.

I have accumulated a significant quantity of flight hours in helicopters due to my professional activity in Power Line Maintenance Inspection and I became a UAS operator to participate in R&D in autonomous flight for the same business. Therefore, I’m interested in the safety topic, both from the manned and unmanned aviation point of view.

Through the discussion and exploration additional remarks arose and these are worth mentioning:

1. Statistics indicated in the Bing response use as calculation base hours of flight and this is the correct standard in the business because flight time is the way to measure the use of resources and operation risk in aircraft and aircrews. However, for humans using aircraft as means of transportation, the relevant metrics is “miles flown” instead of “hours flown”. In specialised businesses such as Power Line Inspection, Search and Rescue and others, hours flown are acknowledged as the key metric.
2. #2 statement by MD “Helicopters also have less margin for error and recovery in case of an emergency, as they cannot glide or divert to a nearby airport like planes can” makes little sense to me and raised an early alarm when I first read the post. Helicopters are supposed to fly under Visual Flight Rules [VFR] with very few exceptions and VFR require pilots to only fly into areas where they can identify a suitable landing spot. Since helicopters can land in tight spaces (from witnessing, I would say “incredibly tight spaces”), helicopter pilots usually have more ease to land than fixed wing aircraft as soon as an emergency landing condition arises. My statement is corroborated by Bing [see Annex 1].
3. In Europe through EASA and in many other countries through their respective National Aviation Authorities, the requirements for reliability grow with the perceived risk of helicopter operations. There are certain applications – such as Power Line Inspections – that require turbine engines that have lower failure rates than piston engines and operations– such the overflight of urban areas – that are restricted to twin engine helicopters so that they can keep in the air during an engine failure. However, there are critical elements in helicopters that forbid redundancy such as main transmission and connection to the tail rotor. This is emphasised in #3 statement by MD.

Annex 1 – Bing (January 3rd 2024)

(I’ve made some minor changes into format for clarity. The content has not been changed).

Hello, this is Bing. According to the web search results, helicopters have a higher crash rate than planes for several reasons, such as:

1. They fly at lower altitudes and encounter more obstacles and hazards.

2. They have more complex and sensitive controls and require more skill and attention from the pilot.

3. They have more moving parts and systems that can malfunction or fail.

4. They perform more frequent takeoffs and landings, which are the most dangerous phases of flight.

5. They are used for riskier operations and missions in challenging environments.

The crash rate for helicopters is 9.84 per 100,000 hours, while the crash rate for planes is 6.84 per 100,000 hours [1]. However, this includes single-engine piston planes that are 10 times more likely to crash than jets. If we compare helicopters with commercial planes, which are the ones we usually travel on, the difference is much larger. Commercial planes have a very low crash rate of 0.06 per 100,000 hours [2]

On the other hand, helicopters have a lower fatality rate than planes, because they can glide and land more safely in emergency situations. The fatality rate for helicopters is 1.3 per 100,000 hours, while the fatality rate for planes is 1.8 per 100,000 hours [1]. You can learn more about the causes and factors of helicopter crashes from this article [2] and this article [3]. I hope this helps. (blush)

Annex 2 – ChatGPT (January 3rd 2024)

(I’ve made some minor changes into format for clarity. The content has not been changed).

Helicopters and airplanes have different designs, capabilities, and purposes, which can contribute to differences in their crash rates. It's important to note that comparing crash rates can be complex, as various factors influence safety statistics. Here are some reasons why helicopters might have a higher crash rate compared to airplanes:

1. Complexity of Operation:

Helicopters are generally more complex to operate than airplanes. They require greater pilot skill and experience due to their ability to hover, take off vertically, and land in confined spaces. The complexity of helicopter operations can contribute to a higher risk of accidents, especially during low-altitude maneuvers.

2. Hovering and Low-Speed Flight:

Helicopters often operate at lower speeds and altitudes than airplanes. Hovering and low-speed flight present unique challenges, and accidents can occur during these phases, particularly if pilots lose control or encounter unexpected environmental conditions.

3. Autorotation and Power Loss:

Helicopters can enter a state called autorotation in the event of an engine failure. While this allows for a controlled descent, it requires skilled pilot intervention to execute safely. Power loss in helicopters poses a significant risk, and if not managed properly, it can lead to accidents.

4. Weather Sensitivity:

Helicopters are generally more sensitive to adverse weather conditions than airplanes. Turbulence, gusty winds, and low visibility can pose greater challenges for helicopters, increasing the likelihood of accidents, especially during critical phases of flight.

5. Mechanical Complexity:

Helicopters have more moving parts and complex mechanical systems compared to airplanes. Mechanical failures or malfunctions can occur, leading to accidents. Regular maintenance and inspections are crucial for helicopter safety.

6. Mission Profiles:

Helicopters often perform missions such as search and rescue, medical evacuation, and law enforcement, which may involve flying in challenging environments. The nature of these missions can expose helicopters to higher risk factors.

7. Training Differences:

Pilots undergo different training for helicopters and airplanes. The skills required for helicopter flight, including managing hover, autorotation, and low-speed maneuvers, are distinct from those needed for fixed-wing aircraft. The learning curve for helicopter pilots can contribute to a higher risk of accidents.

8. Small Helicopter Fleet:

The total number of helicopters in operation is smaller compared to airplanes. A smaller fleet size can result in statistical fluctuations, making it challenging to draw direct comparisons between the two.

It's important to consider that ongoing advancements in technology, training programs, and safety regulations aim to improve the overall safety of both helicopters and airplanes. Additionally, individual accidents can be influenced by a combination of factors, making it crucial to analyze specific incidents to understand the root causes.