Developer Burnout and AI Self-Driving Cars

2236

By Lance Eliot, the AI Trends Insider

Did you realize that apparently more than half of all United States medical doctors are suffering from burnout?

You might at first glance not be overly surprised, since you’ve likely seen how harried most medical doctors are. Often, their patient load is at the max and they barely have time to say hello before they move onto the next patient. Many medical doctors bitterly complain that the nature of the healthcare system prevents them from spending quality time with their patients as they are under strict time guidelines and have little choice in the matter of how they use their time with patients.

You might also be thinking that it doesn’t matter that medical doctors might be suffering burnout. The perception is that they are well-paid anyway, and so if they have to work long and hard hours, so be it. Some might think of them as whiners that don’t realize how good they really have it. This though would be a misunderstanding about the impacts of burnout.

The medical doctors reported that nearly one in ten of them had committed a significant medical mistake, one or more such mistakes or errors, in the three-month period prior to the poll being taken. It is generally well proven already that burnout leads to medical doctors making mistakes or errors, and we now know the alarming frequency in which it can occur.

There can be errors in ascertaining the ailment that a patient has, or maybe a mistake in a prescription issued for a patient, and so on. The burnout therefore can directly and adversely impact the nature and quality of the healthcare provided to patients. In addition, medical doctors can become depressed, have high fatigue, and otherwise be less effective and efficient in performing their medical tasks.

Presumably, burnout is a reversible work-related matter.

If you can detect early enough that someone is suffering from burnout at work, you can potentially provide them guidance on how to alleviate the burnout. Some try stress management techniques to reduce their burnout. Some use the latest in so-called mindfulness training. Some try to seek a balance between the demands of work and their other life pursuits, carving out more time and attention to efforts outside work that enable them to better contend with the work situation.

It is usually unlikely that changes by the individual alone that has exhibited the burnout is going to be sufficient to curtail the burnout. The work situation often needs to also adjust. An organization has to realize what factors are leading to the burnout, and potentially readjust work schedules, or adjust the nature of the work being performed, and so on. Someone that is otherwise well prepared to contend with burnout is still going to have a tough time not getting burned out if the work environment that presumably is causing the burnout does not make adjustments too. It takes two to tango, as they say.

When I mention that burnout is potentially reversible, I’d like to clarify that for some people in some companies it is not reversible. When someone reaches a certain threshold, they can be so far gone that they cannot find within themselves the desire and nor the need to re-commit themselves to work, even if the company offers to try and find a means to do so. I’ve seen some workers that got burned out at a firm and they left in disgust and with no intention of ever returning. That being said, I’ve seen some firms that claimed they wanted to save someone that was burned out, but the firm did nothing more than token attempts to keep the person, which for that person made them even more determined to leave the firm.

When considering medical doctors that get burned out, you need to consider not just the impact on them and their patients, but also take a wider view and consider the larger ramifications. The odds are that if the patient gets less capable medical care due to the burn out, it might also indirectly impact their family and friends. Those family and friends might need to provide other outside care or additional care to make-up for whatever medical errors or omissions occur. The odds are that fellow staff at the medical facility will also suffer, having to either deal day-to-day with a medical doctor that might be difficult to deal with, or need to deal with patients that become irate when they realize they are not getting their desired care. Overall, you could make the claim that medical doctor burnout will raise costs overall for the medical delivery system and all of society accordingly, and also reduce the available medical care for others by needlessly consuming limited available medical resources by the burnout effects.

There are some workers that drive their own burnout. You’ve likely dealt with a workaholic that seems to work all of the time. They say that it makes them stronger and they enjoy it. This can sometimes be true, but more often than not it is the path towards burnout. A workaholic can work themselves to the bone. For some junior managers they think that having a workaholic under them is great, since the person gets so much work done. But, in the end, the person might be a prime candidate for burnout and thus the junior manager has likely done a disservice by not having done something about the matter earlier.

Besides the workaholic, there are other types of workers that can be especially susceptible to burnout. There’s the lone ranger that tries to take on all the work themselves and doesn’t appropriately make use of their fellow team mates. There’s the perfectionist that wants to do everything to the nth degree and often goes overboard in terms of their work. There’s the superhero type that relishes coming to the rescue on efforts and will become overwhelmed with work. There’s the martyr that likes to do tons of work to be able to let others know that they are doing so. Etc.

Besides medical doctors, there are other professions that involve substantial amounts of burnout.

One such occupation are the AI developers that are working on AI self-driving cars.

At the Cybernetic AI Self-Driving Car Institute, we are developing AI software for self-driving cars and besides our own AI developers we also keep in touch with other self-driving car AI developers.

Generally, burn out is pervasive among such AI developers.

Why?

Why Burnout is Pervasive Among Self-Driving Car Developers

You might at first think that it would be an exciting and enjoyable job to be an AI developer for AI self-driving cars. It’s like trying to achieve a moonshot and having been there in the early days of developing the Apollo spacecraft’s that got us to the moon. There’s a thrill about doing something that could change society. It has the potential for great benefits to all of us. That kind of a job should be joyous!

For my article about why AI self-driving cars are like a moonshot, see: https://aitrends.com/selfdrivingcars/self-driving-car-mother-ai-projects-moonshot/

For my article about idealism in AI self-driving car developers: https://aitrends.com/selfdrivingcars/idealism-and-ai-self-driving-cars/

Though it is true that you could perceive the job as striving to achieve new ends and changing our lives, getting there is not all fun and games.

Let’s consider why there is a high chance for burnout for AI developers that are developing AI self-driving cars:

  •         New Ground

AI self-driving cars are pushing the boundaries of what we can do with AI. This is not run-of-the-mill stuff. We are using the latest AI techniques, the latest Machine Learning (ML) capabilities, etc. Many everyday developers often just reuse what they have done before. In this case, it’s new ground with every step you take.

  •         Life-or-Death System

It’s one thing if you are pushing the boundaries of AI for let’s say a financial system, but in the case of a self-driving car, it’s a life-or-death matter. If your software hiccups at the wrong time, it could mean that the car will hurtle itself into a wall and kill the human occupants. Or, it could swerve unintentionally and kill pedestrians. And so on. This is serious stuff.

  •         Real-time System

Whenever you are developing software for a real-time system, it tends to increase the difficulty factor. One of the first real-time systems I was involved in, years ago, involved a real-time controller for a roller coaster. I can tell you that we sweated quite a bit about how to get the timing just right and make sure the software was always able to handle whatever happened in real-time.

  •         Intense Pressure

The pressure by the auto makers and tech firms to get their self-driving cars on the roadways is intense. Every day you see new announcements about one self-driving car maker is going to get to the market sooner than the other. This kind of gamesmanship is often taking place without regard to what the actual AI developers can do – it’s about what they leaders are telling the marketplace. Deadlines aplenty. Irrational deadlines aplenty.

  •         Lack of Specs

Many of the auto makers and tech firms are developing their self-driving cars on-the-fly in an agile method and doing so without a definitive set of specs. To some degree, it’s make it up as you go. Some of the ideas that are being delved into these projects are dreams rather than something that can be actually achieved. AI developers are often told, rather than asked, what can be done.

  •         Spotty Peer Expertise

There aren’t many that have industrial style expertise in developing software for cars, let alone for AI self-driving cars. Thus, it is somewhat unlikely that an AI developer can depend upon their fellow AI developer on their team to lend a hand. The odds are that they are all mainly in-the-dark and trying to figure out things as they go along.

  •         Highly Secretive

The self-driving car efforts by each auto maker and tech firm are typically being done in a skunk works operation that’s considered for secure eyes only. This secretive manner makes sense because everyone is trying to do their own thing and they don’t want others to steal it. But, this also makes things harder for the AI developers since it narrows whom else they can turn to for assistance. In many cases, they aren’t supposed to talk about their work with family and friends – it’s like being in the CIA.

  •         Shift From R&D

For many of the AI developers in the self-driving cars field, they most recently were working at an AI research lab at a university. That’s a whole different kind of work environment than in industry. For example, at a university, there is often the view that failing on something is OK since you are doing experimentation and not everything will work out. No matter what you hear about Silicon Valley saying to fail first and fail fast, I assure you that with the pressures to get self-driving cars going, the “let’s try failing” model is verboten.

  •         Long Hours

With the vast amount of work to be done for the AI of a self-driving car, there are long hours involved. It can be frustrating too because as mentioned it is punctuated with trying new things and hoping they will work. And, you can’t readily explain to family and friends why you are working late and on the weekends, other than they know vaguely you are working on something important and hush-hush.

  •         Other

There are a myriad of other factors involved too. For example, even the tools used to develop the AI systems are at times brittle and still untried. It would be like trying to make a house and you have hammers and screwdrivers that no one knows for sure will work properly.

Now, I realize that many of these AI developers are getting paid big bucks. As such, similar to the perception about medical doctors, you might have little sympathy about these AI developers possibly getting burned out. You might say they should relish their moment in the sun. Now’s the time to make enough bucks to then retire.

Well, maybe, but let’s also consider the impacts of burnout, similar to the concerns when medical doctors experience burnout.

In the case of the AI self-driving cars, it can lead to the AI developers making errors or mistakes, more likely than they might have otherwise. Perhaps mistakes are made in the machine learning and so the AI system is unable to properly interpret a road sign. Or, perhaps there’s a bug in the code that when the self-driving car reaches a particular speed that the code burps and gets stuck in a loop that it can’t get out of.

Here’s the major actions that an AI self-driving car undertakes:

  •         Sensor data collection and interpretation
  •         Sensor fusion
  •         Virtual World model updating
  •         AI Action Plan updating
  •         Cars control commands issuance

See my framework about AI self-driving cars:

https://aitrends.com/selfdrivingcars/framework-ai-self-driving-driverless-cars-big-picture/

Impacts of Burned-Out AI Self-Driving Car Developers

A burned out AI developer can be “lazy” when it comes to testing and decide that they’ve done too much testing already. Or, they might have an attitude of “why test it” since they don’t believe the whole thing will work anyway. Or, they might fix something that they find as broken, and in the effort of rendering the fix they inadvertently and unknowingly introduce another problem into the code.

Of course, any of these aspects can happen to any developer. And they do. But, as mentioned earlier, it is magnified in the case of AI developers for self-driving cars due to the pressures involved, and the pushing of new boundaries, and so on. Plus, this is a real-time system that involves life-and-death aspects. Thus, this happening for AI developers of self-driving cars has especially important and significant ramifications.

Imagine the problems of AI code that is half-hearted and does the sensor data collection interpretation. Or that does the sensor fusion. Or that does the virtual world modeling. Or that does the AI action plan updating. Or that does the cars control commands issuance. What also can hamper things is that an error in one of those crucial components can compound itself by then misleading the other components. It can have an adverse cascading impact. This includes the potential for the Freezing Robot Problem.

See my article about the dangers of the Freezing Robot Problem in AI self-driving cars: https://aitrends.com/ai-insider/freezing-robot-problem-and-ai-self-driving-cars/

What can be done about the burnout of AI developers that are creating the next generation of self-driving cars?

First, it’s vital to acknowledge that the burnout can and does exist. There are some firms that are blind to burnout and don’t know it happens. They often just say that Joe or Samantha need to take a day off, and when they get back they’ll be fine again. This kind of Band-Aid approach fails to recognize the depth and seriousness of true burnout, and the lengthy and complex process to typically undo it.

Next, watch out for the burnout culture that some firms seem to foster. I say this because there are many Silicon Valley firms that actually tout their burnout rates. They like to chew-up people. They make it into a macho kind of atmosphere and try to project an image that only the strong survive. In their viewpoint, if you aren’t already on the verge of burnout, get there or get out. I’ve been waiting to see what happens when those employees so treated decide to finally lawyer-up and go after those firms for the cruddy work environment. We’ll see.

In some firms, they buy a ping pong table or a foosball table and that’s their way of telling the employees to not burnout. Somehow, you are supposed to take time off from your non-stop high-pressure AI development work, and by playing ping pong a few times a day you’ll not get burned out. Doubtful.

Firms that are serious about detecting, mitigating, and preventing burnout will go out of their way to try and arrange the nature of the work and the work situations to deal with burnout. They need to hire the right people, put in place the right managers, provide the right kind of leadership, and otherwise aim to gauge how much work can be reasonably done and by whom. There are many key decisions being made about self-driving car designs, and the coding, which will either aid the fulfillment of self-driving cars, or will have the opposite impact and completely undermine the advent of self-driving cars.

See my article forewarning the groupthink occurring about AI self-driving car development: https://aitrends.com/selfdrivingcars/groupthink-dilemmas-for-developing-ai-self-driving-cars/

Some say that burnout in the workplace leads to the erosion of the soul. I know one promising AI developer that nearly collapsed at the pace that he was going, and decided that it was just too unhealthy to continue on this kind of work. He’s since switched to another occupation entirely. That’s a shame as we already have too few well-qualified AI developers to start with. And we need a lot more of them to ramp up for achieving a true Level 5 self-driving car.

I’ve seen some AI developers that have emotional exhaustion from work burnout. One that went home and took it out on his wife and kids. Another one that became so cynical that he pretty much was approving any kind of code going into production. He had lost the belief in caring for his work. Now, you might say that there should be double-checks to catch these kinds of things in terms of faulty designs and faulty code, but with the go-go atmosphere and high pressures to produce, there are developers that look the other way and figure that it’s up to the other person to make sure their stuff works properly.

I had mentioned earlier that if a medical doctor makes mistakes due to burnout, the patient suffers and also so do a lot of other stakeholders. The same can be said of the AI developers for self-driving cars. They can each in their own way lead to self-driving cars that just aren’t ready for prime time. Unfortunately, those at the upper levels of an auto maker or tech firm might not know or care to know, and just insist that the self-driving car be put onto our roadways. If those self-driving cars harm humans, it’s bad and it will also produce a backlash against self-driving cars overall.

And if that happens, if we have haywire AI that stop or stunts the advent of self-driving cars, it undermines or delays the potential benefits to society that we’re all hoping that self-driving cars will derive. Thus, in that manner, even just one burnout can be like the butterfly that flaps its wings on one side of the earth and it ultimately leads to being felt on the other side of the globe.

I implore the auto makers and tech firms to carefully do an assessment of how they are treating their AI developers, and if it’s “burnout city” then they would be wise to step-back, take another look, and see what can be done to overcome it. All of us need to watch out for that last straw on the camel’s back that will break the spirit of our most prized workers, those AI developers, tasked with creating the future of society via the advent of self-driving cars.

Copyright 2018 Dr. Lance Eliot

This content is originally posted on AI Trends.