Dr. Lance B. Eliot, AI Insider
Wanted, rockstar AI developer. You see this in numerous job ads these days (some claim that it should be stated as two separate words, namely as “rock star” and that if you use “rockstar” you need to capitalize it to be “Rockstar” since this refers to a particular brand or trademark; confusing!).
Anyway, when you see such job ads, the question arises as to what the hiring firm and the hiring manager really are seeking to find.
It could be that the firm has no idea what a rockstar AI developer is or does. Instead, the company is hopeful of making the position itself seem more impressive, and a handy way to do so is to pump-up the volume surrounding the job ad.
Or, it could be that the firm has never hired an AI rockstar and in this case they are hoping, no say begging, or make it beseech that they must have in their midst an AI rockstar developer.
In talking with various firms that have been looking for an AI rockstar developer, which I often get asked for recommendations, I often find that the hiring manager didn’t actually put into the job requisition that they were seeking a rockstar. They indicated in flatter and less flashy terms the nature of the position requirements. The HR (Human Resources) team, sometimes nowadays referred to as the Talent Management group, opted to add the rockstar indication when they posted the position for the world to see.
By planting the AI rockstar proclamation as a flag of announcement in a job ad, HR hopes that it will discourage the AI unwashed, particularly those that decide they want to apply for the job and yet aren’t up-to-par by the perceived AI stupendous standards of the HR team.
Admittedly, getting a true AI rockstar can be quite advantageous.
First, it allows the firm to tout the kind of talent they attract. Look at us, it says, given the fierce competition for the top 1% of AI developers, we got one. This is especially the gambit of the smaller firms or startups. They know that the bigger tech firms can readily attract these top enders.
Secondly, in my experience as a manager and leader, I’ve found that there is absolutely a difference in capability, effectiveness, and productivity between an average AI developer and an outstanding AI developer.
I am not knocking the average AI developers. They are good to have. They get their jobs done. They grind out the work. Yet, if I could have at least one or more of the outstanding AI developers mixed into the pack, it would be huge game changer for what the AI team can accomplish.
Don’t be fooled though into thinking that if you toss one outstanding AI developer into a group of everyday AI developers that magic will somehow materialize. Not so.
There is a solid chance that some of the everyday AI developers will resent the outstanding AI developer.
This in my book comes down to the lot of the AI manager or leader. If they aren’t doing their job properly, they are going to make a mess of the AI talent that is amassed, likely whether there are any AI rockstars or not (and, worse so when there is an AI rockstar, since the odds are the rockstar will not be leveraged appropriately).
AI Rockstar Often Not an Effective People Manager
The problem often times is that the AI rockstar is not a manager. This kind of makes sense because they are usually being hired as an individual contributor, a heads-down developer. They are not being hired to be a manager. Thus, the actual AI manager was handing over the AI project keys to someone that wasn’t able to manage (of course, apparently neither was the anointed manager!).
I’m not saying that you cannot be a highly technical AI developer and also be a manager. What makes the issue confounding is when a firm that is doing the hiring hasn’t figured out what they want or need.
Ideally, if you are looking for an AI rockstar developer that is purely a developer, the position actually matches to that need, and furthermore there is an AI team manager that knows how to properly make use of that talent.
Jerk-Rockstar Causality Confusion
This brings up one aspect that I admit gets my goat whenever it arises. I call it the jerk-rockstar causality confusion.
Allow me to give you an example. One firm was trying to hire an AI rockstar and they brought me into the hiring process to help vet candidates.
They already had two finalists. One of the finalists was a real jerk. You could discern this the moment you spoke to the person. They were full of themselves, they acted like they could part the sea, and when I asked some fundamental AI questions as means to gauge what they really knew, the person immediately rejected them as beneath them and refused to answer the questions. The other candidate was more moderate and nearly reasonable, especially in contrast to the complete jerk candidate.
After I spoke with the two finalists, I went over and chatted with HR. The recruiter on the HR team said that the candidate that was “at times difficult” was clearly the better candidate and wanted to know what I thought. I asked how the recruiter assessed that the “at times difficult” candidate (i.e., the jerk) was the better of the two candidates?
Answer: Because AI rockstars are jerks and since the one candidate was a jerk, it meant the candidate was better than the other candidate (the one that was not so much of a jerk).
Say what? Apparently, there is a causality that if you are a jerk, ergo you are an AI rockstar. Likewise, presumably, if you are not a jerk, ergo you are not an AI rockstar.
Some AI rockstars have actually figured out this mindset exists and therefore they ramp-up the jerk factor when they meet people or are trying to get hired.
In any case, I mention this gets my goat because there are AI rockstar developers that are not jerks.
As a quick recap on this diatribe about AI rockstars: they do exist, they are worth their weight in silver, they are not necessarily jerks, they can enhance an AI team, they can bring a glow to a firm that lands them, the “title” can be an attractor for a rockstar that wants to be recognized for what they can bring to the table, it can be a handy means of weening out the non-rockstars, and it can be harder than it might seem to discern whether someone is a true AI rockstar.
AI Self-Driving Cars Need Their AI Rockstars Too
What does this have to do with AI self-driving cars?
At the Cybernetic AI Self-Driving Car Institute, we are developing AI software for self-driving cars. We and everyone else involved in AI self-driving cars such as the auto makers and various tech firms are all on the hunt for AI rockstar developers, plus we are often asked to aid in the identification, selection, hiring, and onboarding process.
Allow me to elaborate how this pertains to AI self-driving car efforts.
I’d like to first clarify and introduce the notion that there are varying levels of AI self-driving cars. The topmost level is considered Level 5. A Level 5 self-driving car is one that is being driven by the AI and there is no human driver involved.
For self-driving cars less than a Level 5, there must be a human driver present in the car. The human driver is currently considered the responsible party for the acts of the car. The AI and the human driver are co-sharing the driving task.
Another key aspect of AI self-driving cars is that they will be driving on our roadways in the midst of human driven cars too.
Returning to the topic of AI rockstar developers, let’s consider how these rockstars are manifested in the context of the AI self-driving car industry.
Where to Find an AI Rockstar?
If you were looking for an AI rockstar developer that has AI self-driving car expertise, where would you look? By-and-large, many of these AI developers have been sourced out of various university research programs that have had a focus on autonomous vehicles.
Indeed, many of the key AI self-driving car specialists and leaders of today that are at the major auto makers and tech firms developing AI self-driving cars came out of the DARPA (Defense Advanced Research Projects Agency) Grand Challenges that took place in the early 2000s.
In 2004, the first of the DARPA Grand Challenges took place, involving a race in the Mojave Desert that generally parallels Interstate 15 in California and required trying to make a 150-mile journey with an autonomous vehicle in the desert. Though none of the vehicles were able to successfully complete the race, and did not win the coveted $1 million cash prize, this effort approved by the U.S. Congress was able to kick-up further interest in creating AI self-driving cars.
Then in 2005, there were five winning autonomous vehicles, arriving at the finish line in this order, and for which I indicate the name of the vehicle and who provided it: (1) Stanley of the Stanford Racing Team, (2) Sandstorm of the Red Team from CMU, (3) H1hglander of the Red Team from CMU, (4) Kat-5 of the Team Gray from The Gray Insurance Company, and (5) TerraMax of the Team TerraMax of the Oshkosh Truck Corporation.
For 2007, the third DARPA Grand Challenge took place in a closed-track urban-like setup setting and became famously known as the “Urban Challenge.” This took place at an Air Force Base in Victorville, California. Six of the submitted autonomous vehicles were able to complete the course, which included having to abide by various traffic related rules that has been stated.
That was over a dozen years ago. When the rush toward AI self-driving cars began just a few years ago commercially, many of those that had been directly or indirectly involved in the DARPA Grand Challenges either flocked to private industry or were lured out of universities into commercial enterprises.
But there are only so many of those such AI developers and it is insufficient as a pool for the number needed to fully resource the numerous and varied AI self-driving car efforts underway. As a result, there are now some being lured out of university research programs that were not around during the days of the DARPA Grand Challenge. These are researchers that came along after those days.
We are also on the verge of seeing poaching or the musical chairs game of AI developers hopping from one AI self-driving car effort to another. This hasn’t happened too much just yet, partially due to the aspect that many of these efforts are still relatively new. There is also the sometimes golden handcuffs that firms use to try and keep their AI developers from jumping ship to another firm. I’ve predicted that we’ll soon see more and more of movement between the AI self-driving car efforts.
Real-Time Systems Experience Needed To Develop for AI Self-Driving Cars
We’ve been helping to train those that are somewhat versed overall in AI about the aspects of AI self-driving cars, though it is a steep hill to climb if the person doesn’t have already some kind of relatively in-depth real-time systems experience.
AI self-driving cars and their systems all work in a very tight time constrained setting and involve the core aspects of real-time systems, plus these are real-time systems involving multi-ton cars that can bring about life-or-death.
In aiding the interviewing and selection process for some of the AI self-driving car searches for AI rockstars, one aspect that repeatedly comes across is the often-seen dogmatic perspective. It is somewhat common that a person versed in AI and self-driving cars or autonomous vehicles might have a particular bend or strongly wedded approach or technology that they adhere to.
This is akin to a computer programmer that insists on using a particular programming language and refuses to consider any other coding languages. Or, one that insists on using a particular software package or has opted to place all their eggs into one basket, and knows only that particular package, therefore they claim that it is the only and best way to go.
I recall one potential AI rockstar candidate that insisted cameras were the best way to collect sensory data for an AI self-driving car and eschewed the use of LIDAR. This candidate was absolutely convinced that LIDAR was overly expensive and not worth the effort to include in an AI self-driving car. Sidenote, I refer to this as the “myopic” or cyclops view of AI self-driving cars, wherein a person believes there is only one way that things are to be done.
What was especially interesting, and revealing was that the person had spent their entire prior efforts on traditional vision processing involving cameras. They had put at most a token effort towards learning about LIDAR. In that case, this “expert” had really had little basis for offering such a strong opinion of the tradeoffs between the two.
I also tried to point out that there is a rapid pace at which LIDAR costs are coming down, simultaneously the accuracy and features are rapidly increasing.
I also pointed out that this seemed to be a potentially false “mutually exclusive” type of debate. Does one necessarily need to choose between using cameras versus using LIDAR? Most of the AI self-driving car efforts to-date are using both (though, Tesla is a notable exception, and I’ve warned many times that I believe Elon Musk’s claim that LIDAR won’t be needed is sadly misguided and he’ll eventually regret the choice, which he has even stated might turnout to be the case).
This potential AI rockstar was surprised to be challenged on this point about the cameras versus LIDAR matter. To-date, he had been able to browbeat most interviewers into submission by using arcane jargon and attempting to bolster his argument about cameras, which really was an argument about why he should be hired. I’m not saying that he lied, and I do believe he sincerely believed in the cameras approach, but I am saying that his lack of awareness and coupled with his personal bias is what led to his insistence.
I would also say this was another example of the jerk-rockstar causality confusion. Those that had interviewed the candidate liked the sense of confidence and spirit of the person and though there was an underlying know-it-all and jerk factor, this merely added to the person’s glow that they must be an AI rockstar.
When I spoke with some of the executives and the hiring manager, I emphasized that since the firm was already using LIDAR, they were going to be bringing into their midst someone that was adamantly opposed to this part of their strategy and efforts. I was told that they would just keep the person in the cameras and vision processing team. No need to have them deal with the LIDAR team.
Sigh. I tried to point out that if they wanted to ensure a war between their teams, they certainly could so proceed. At every turn, this person would likely try to undermine the other team. The other team would likely become antagonistic toward the cameras team and if there wasn’t bad blood yet, the company would soon be bathed in bad blood. This didn’t seem to be a smart way to try and seamlessly craft an AI systems for which all of the parts need to work coherently and cohesively.
This does bring up another facet about AI rockstar developers in the AI self-driving car niche.
Typically, these AI rockstars have a specific and narrow area of skills and technology attention. That’s fine, as long as this is realized and leveraged. Someone that is highly skilled at the sensors part of self-driving cars might not be familiar with the car controls aspects. Thus, you cannot just slam dunk someone into the car controls side if you’ve hired them based on their sensor-focused skills.
I also forewarn to be on the watch for a kind of technological bigotry, such as the candidate that was so convinced that cameras rule the world. It’s handy for someone to be passionate about their area of expertise, but it is another thing when they bash another area of technology and want to fight against it. That’s where you are bound to have problems arise and it will likely shakeup any AI team.
I’ll cover a few other salient points about potential AI rockstars and self-driving cars.
One point is that they sometimes are so strongly opinionated that when they get onboarded and have a chance to look under-the-hood of the AI systems being developed, they can suddenly decide that everything is wrong and that there should be a do-over.
That can be quite a shock to the firm.
If the AI rockstar is actually right and they have discovered that there are serious and severe flaws, well, okay, thankfully they have found this, preferably before the firm has gotten too far along on their AI self-driving car efforts. On the other hand, if they person is being a jerk and merely spouting out false failings, maybe to boost their own sense of importance, or perhaps based on a misjudging of what they’ve found, it is going to likely cause chaos.
Imagine a firm that has invested perhaps millions upon millions of dollars into their AI self-driving car development, along with multitudes of expensive AI developer time and effort, and have someone that walks in the door and proclaims that it is a waste. Yikes! The newly hired AI rockstar will likely get heard because it is assumed that they are an AI rockstar, since they got hired under those auspices, and so it will be difficult to quiet down such a charge.
This could take the firm in a path that will last for weeks or months of internal handwringing and debate. All of which might be warranted, or might be a false “the sky is falling” and that drains the attention and monies of the firm. With the frenetic pace of AI self-driving car efforts, and the desire to keep ahead of the other AI self-driving car firms, getting bogged down in an acrimonious internal debate that perhaps has no basis will be draining and likely cause the firm to fall behind.
I don’t want to suggest that the AI rockstar is necessarily wrong if they do find problems. In fact, it could be that others within the AI team have had qualms about the AI system, but they were either unsure of how to voice those qualms or felt they would not be heard if they did. Internal naysayers are often cast aside and gain little by coming forth, other than a personal sense of doing what they believe to be right. When a newly hired AI rockstar opts to rock-the-boat, it can be an opportunity for suppressed members of the AI team to voice their concerns.
Everyone wants to hire a rockstar. And, why not? The implication is that if you are hiring someone less than a rockstar, you are presumably settling for the mediocre. What firm wants to run job ads saying they want to hire mediocre AI developers? Imagine the reaction. It would hurt the firm’s reputation in the industry and it would likely have the internal AI teams feel like they have also been slapped in the face.
What exactly constitutes being an AI rockstar? Is it the number of years’ experience in developing AI systems? Is it the kind of AI systems developed? Is it the prominence among your AI peers and within the AI community? In one sense, an AI rockstar status is in the eye of the beholder.
Some believe that AI rockstar labeling might be gradually running thin. Critics would say that it is akin to how school children are all told they are winners when playing a sport, even though they were participants and did not necessarily place in the top three or top five positions. Maybe everyone that does AI wants to believe they are an all-out super-duper AI rockstar. In that case, the search for AI rockstars is going to be easy, pick anyone that knows AI.
In any case, I advise firms to be mindful that when hiring an AI rockstar they ought to be carefully considering how the person will fit into their existing AI team. Furthermore, there needs to be a solid AI manager overseeing and managing the AI team, otherwise adding the AI rockstar could inadvertently cause the AI team to get waylaid or go kilter. It takes a village to make an AI self-driving car and hiring one AI rockstar as a solo act is not going to get you there.
For free podcast of this story, visit: http://ai-selfdriving-cars.libsyn.com/website
The podcasts are also available on Spotify, iTunes, iHeartRadio, etc.
More info about AI self-driving cars, see: www.ai-selfdriving-cars.guru
To follow Lance Eliot on Twitter: @LanceEliot
Copyright 2018 Dr. Lance Eliot