Kinship Of AI Systems Leads To AI Hidden Alliances, Emerging In Advent Of Autonomous Cars

Dr. Lance Eliot, AI Insider

Image for post
Image for post
Blood is thicker than water, so it is said, and AI kinships might be similar

If you look at history there is a lot of credence to the power and attraction of kinship. For those of you that have been Game of Thrones fans, kinship was certainly a front-and-center theme throughout the entire series, though admittedly a fictional portrayal, but I dare say we might all agree that history showcases the same phenomena.

If you are a history buff, you likely know that the famous Heidelberg manuscript from the 13th century indicated that kin-blood is not spoiled by water. In more modern times, we’ve come to express this as the now-classic saying that blood is thicker than water. Whichever way you might prefer to state it, the underlying notion is that bloodline family and familial relationships are considered a very strong bond.

Indeed, some would assert that the familial bonds are stronger than any other kind of friendship or relationship that you might ever formulate. Family and bloodline prevails over anything else, in their view.

Humans have often had to make tough choices in life and at times seemed to choose the bloodline, even when it might have been more sensible to not do so. Animals seem to also have a familial tendency and you can readily watch online videos of wild animals that will take great chances to save or protect their own bloodline offspring. Kinship appears to be quite pervasive.

It is difficult to clearly say why this kinship aspect matters so much.

Those that study the nature of evolution and abide by Darwin’s theories would suggest that it is rather apparent why the bloodline would be so revered. When you are faced with the basics of survival, you need to have others that can protect your back. By becoming a kind of pack, your chances of survival are presumably enhanced. The question then becomes whom can you or should you form a pack with? In a rudimentary caveman or cavewoman manner, it would have made sense to focus on your own bloodline.

One aspect is that your own bloodline would be a known more so than an unknown. Those with your bloodline were more likely to have similar characteristics as you. They would tend toward the same physical attributes and presumably similar cognitive and personality attributes. This would appear to make the connectivity of the pack a more likely magnetic or attracting element and promote cohesiveness. Strangers would be less likely to stick out their neck to help you, while your own type would perhaps be more willing to go the extra mile for your sake.

AI Autonomous Cars And Kinship

What does this have to do with AI self-driving driverless autonomous cars?

At the Cybernetic AI Self-Driving Car Institute, we are developing AI software for self-driving cars. One interesting aspect will be how various AI self-driving cars act toward each other, for which there might be a kinship element involved.

Allow me to elaborate.

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.

Driving And Kinship Versus Non-Kinship With Other Drivers

Suppose you are driving your car on a busy highway. Your daughter is driving her car and is just ahead of you in traffic. You catch-up with her and realize that she is going to try and make a right turn into an upcoming shopping center.

The highway traffic is moving at a fast clip and you realize that when she tries to make the right turn, it will necessitate her potentially having to make the right turn very quickly so as to not slowdown traffic. Making such a turn rapidly might not be a good idea and could lead to her potentially hitting a pedestrian or another car that is near to the entrance into the shopping center where she is going to make her turn. You are also worried that when she starts to make the right turn, a car from behind her on the highway might be impatient and fail to notice that she is going to make a turn and might ram into her car.

Therefore, you decide to stay directly behind her and when she approaches the point of the turn, you opt to especially slowdown, pump your brakes, and otherwise try to block the traffic in your lane. This is a means of trying to help her make the turn. It is almost like playing football, wherein one player blocks others to allow the quarterback to make their move. You are creating a kind of traffic buffer so that she can make the right turn at a more measured pace and to prevent other highway traffic from potentially pressuring her or possibly even hitting her car.

What a great parent you are!

I’d like to now change the scenario just a tad. Suppose the same situation arises, but this time the driver in the car ahead of you is unknown to you. As far as you know, the driver is a complete stranger. How might that alter your actions?

The odds are that you would not be quite so accommodating to a driver that you don’t know.

We might say that your kinship shaped your driving behavior.

In other words, if you know the person that is driving a car that is near you, you might well change how you drive, depending upon what your “relationship” might be associated with that person. I think we can all agree this seems plausible. I would even say it is more than plausible, it is probable that you would change your driving behavior.

I realize there are some sticklers out there that will claim they always drive the same way. They are always courteous and fair to other drivers. Always. Or, maybe they are always a jerk to other drivers and won’t change their driving behavior for anyone. They don’t care who is in the other cars. By gosh, they are going to drive as they drive, all the time, the way they do, and continue to cut-off other drivers and treat them like dirt.

I’d bet that most of us do change our driving behavior depending upon whether we know the driver that is in another car nearby us. Let’s agree to this notion.

Driving Behaviors Dependent Upon The Known Or Unknown Other Drivers

We’ll establish then that your driving behavior can be impacted based on whom else is driving another car and for which your driving and your car will in some manner interact with that other car.

I’m not suggesting you will always suddenly become a kinder and gentler driver to accommodate the other person, certainly there are exceptions.

By the way, notice that each of the situations depicted so far have been undertaken without any direct communication between you and the other driver.

Your driving behavior, in of itself, might have been your form of communication.

I’d claim that we have these situations continually during our driving efforts. We have situations arise, and at times another driver might be intentionally helping you out, while in other instances the other driver’s behavior just so happens to help you out. You might not ever know what was in the mind of the other driver.

Sure, we sometimes get a hefty clue of what the other driver has in mind.

Our driving behavior can alter depending upon the other driver. If the other driver is someone that we know, this is a likely behavior changing factor. We might directly communicate with the other driver, or we might not. Our behavior alone might be our form of communication and no other kind of “direct” communication is undertaken.

We spontaneously collaborate with other drivers when we drive. There is not necessarily any prearranged agreement between one driver and another. There is a lot of discretion when you are driving a car. Of course, you are supposed to drive lawfully, but within the legal definitions for proper driving there is a great deal of latitude about driving.

Autonomous Cars And AI Knowing Or Not Knowing Of Other Drivers

Consider what is going to happen once we have true AI self-driving driverless autonomous cars on our roadways.

First, realize that not all AI self-driving cars will be the same.

I emphasize this due to the commonly false notion that all AI self-driving cars will be the same. When I give presentations at conferences, I often find that people assume that the AI of self-driving cars will be entirely the same AI, no matter who developed the AI and want kind of car you are driving.

Wrong!

Let’s set the record straight. The AI of the self-driving car that is made by auto maker X or tech firm Y will be different from the AI of the self-driving car maker Z or Q. There is no underground secret agreement between these auto makers and tech firms. They are all pursuing the AI in their own proprietary ways. There might be some amount of open source in their AI systems, providing a bit of intersection, but otherwise each will be having its own idiosyncratic AI.

There is also no set standard that specifies exactly how the AI is supposed to be built and nor how it is supposed to act. Similar to how humans are able to have latitude when driving a car, this lack of any enforced specification means that the AI of each of the auto makers and tech firms can differ from that of the AI of each other in terms of how the AI drives a self-driving car.

You might readily have one AI system from say auto maker X that is a very cautious driving system. It avoids taking any chances while driving. Imagine the teenage novice driver that you sometimes get upset about when you get behind them. That could be the AI system of auto maker X, in the sense of exhibiting driving behaviors of a similar nature to the novice teenage driver.

Meanwhile, auto maker Y has gone a different path. For their AI, they decide to make it the type of driver that goes all out. It will readily be the first one to zip ahead when the light turns green at a stopped intersection. When it makes turns, they are always done with gusto. It exhibits completely different driving behaviors than does the AI system of auto maker X.

I realize you might object and claim that isn’t it possible for auto maker X to potentially “copy” the driving behaviors that are being exhibited by auto maker Y’s AI? Or, the other way around? Yes, it is. But, keep in mind that if you are trying to force both of their AI’s to become the same, I wonder why you today tolerate the aspect that there are different kinds of cars that you can buy, some that are high-powered, some that are stylish, etc.

In essence, I am suggesting that we are likely to have AI systems for self-driving cars that abide by the legal elements of driving and yet differ in other driving behavior respects. It could be that each of these auto makers and tech firms eventually decide to “copy” each other and have the same set of identical driving behaviors, though I’ve questioned this notion and tried to explain why I believe that the “commoditization” of AI self-driving cars is unlikely to occur.

Puzzle Pieces Fitting Together

We are now finally ready herein to piece together all the pieces of this puzzle. What is the puzzle?

We are going to have AI self-driving cars on our roadways and I hope you now agree with me that those AI systems will differ, meaning that the AI driving behaviors will also differ.

Presumably, the AI self-driving cars of auto maker X will exhibit the same driving behaviors as those of the other AI self-driving cars of auto maker X, namely, there will be a kind of kinship. There, I said it, the word kinship has now arisen.

The AI of the auto maker X will be the same and drive the same way, and we could potentially even pick out of a line-up which AI is being used by a particular “anonymous” self-driving car.

How so? If we disguised a bunch of AI self-driving cars so that we could not recognize the car maker, and we put them into a test track someplace and had those true Level 5 AI self-driving cars drive around, I dare say that based on what they do, we could tell you whose AI it was.

I’ll use my somewhat extreme example from my earlier point about auto maker X and auto maker Y. During a run in the test track, we observe that one of the self-driving cars is always being overly cautious. The other one tends to butt up against other cars and tries to jam itself into a lane. I think it would be relatively easy to guess that the cautious self-driving car was using the AI of auto maker X, while the more aggressive self-driving car was using the AI of auto maker Y.

Again, don’t discard my claims due to the extremes of suggesting one is overly cautious and the other is overly risky. There are a lot of other capabilities that will differentiate one AI self-driving system from another in terms of driving behaviors. I am only using the extremes herein to illuminate my point.

Heterogeneous AI Systems And Differing Behaviors

We’ve now nearly got all the pieces of the puzzle in place.

With the auto maker X and its AI that exhibits some set of driving behaviors, and with the auto maker Y and its AI that exhibits some other set of driving behaviors, we now have the potential for a kind of kinship.

The AI of X might have been developed under the assumption that other self-driving cars that are also using the AI of X will act in certain driving ways.

For other self-driving cars and their variants of however they’ve done their AI, the auto maker X is assuming those other cars are being driven by however they are being driven. This implies that the non-X AI’s are considered the same as human driving behaviors and whatever happens to appear while on the road is just how that driving is taking place.

Let’s imagine that the auto maker X has made their AI to be this sweet and kindly self-driving driver. It comes along on a highway and detects that the car ahead of it wants to make a right turn into a shopping center. Since this AI is the nice-driver, it opts to slow down and create a traffic buffer for the other car that is turning into the shopping center.

But it might be that the auto maker X’s AI does not always necessarily undertake that approach. Suppose the car ahead did not signal and did not provide any kind of in-advance indication that it was going to make that right turn. The AI of the self-driving car had no way to realize that it could help the other car. Therefore, the car ahead perhaps makes the right turn, doing so without any warning, and the AI of the self-driving car was unable to assist, not having had any means to gauge what might happen.

On the other hand, suppose that the auto maker X’s AI, being as courteous as it is, if it was making the right turn, it would have turned on its turn signal well in-advance of the turn, and perhaps lightly touched its brakes, trying to convey to any cars behind it that it was wanting to make a right turn.

Revisit the scenario. Let’s pretend we have the car making the right turn and it is a self-driving car by auto maker X and running their AI. Let’s further pretend that the car behind it is also a self-driving car, and it perchance happens to be a self-driving car also by auto maker X and running their AI. The self-driving car making the right turn is courteous and forewarns traffic, and the car behind it is courteous and upon detecting that the right turn is desired the AI opts to aid in doing so.

Kinship!

You could argue that the AI of the two self-driving cars, being of the same “bloodline” and acting in the same driving behaviors, have aided each other. You might say they did so out of kinship of each other. It is akin to you driving your car behind your daughter and opting to aid her in making her right turn. Isn’t love grand.

There wasn’t any direct communication per se about this collaboration. It was merely predicated on their native behaviors. It could be like two plants of the same bloodline, each aiding the other, doing so not necessarily because they talked about it, but due to their inherent embodied natures that click with each other.

How’s that for a plot twist!

Adding Communication Into The Matter

We can now embellish the kinship aspects. I’ll add communication into the equation.

AI self-driving cars are going to be outfitted with V2V (vehicle-to-vehicle) electronic communications. This will allow AI self-driving cars to communicate with each other. You might have an AI self-driving car driving down the road and it encounters debris. This AI self-driving car sends out a broadcast to alert any other nearby AI self-driving cars that there is debris in the roadway. The other AI self-driving cars that are nearby receive the message and they switch lanes to avoid coming upon and hitting the debris.

A self-driving car trying to make aright turn into the shopping center could broadcast via V2V a message letting other nearby AI self-driving cars know that the right turn is coming up. The other AI self-driving cars would presumably receive the message and ascertain what they will do about it.

I’m sure that you are already assuming that of course the other AI’s will all be courteous and upon this V2V notification slow down and make sure that the right turn can be made safely. Why do you believe that to be the case? Because you have fallen again into the trap that all AI’s of the all different self-driving car makers will be the same. As mentioned, they won’t be.

Perhaps auto maker Y’s self-driving car happens to be behind the AI self-driving car of auto maker X that is trying to make the right turn. The AI of Y might decide to avoid the right turn situation entirely and instead switches lanes. That’s perfectly legal. There was no legal requirement that the auto maker Y self-driving car has to perform a traffic buffer block for the right turning car. That’s not written in stone someplace.

Therefore, the addition of communication does not necessarily take us out of the kinship mode. You could still claim that kinship will make a difference. It makes a difference because of the driving behaviors and how the AI of the same self-driving cars will be of the same “bloodline” and potentially therefore be more likely to play well together.

Auto maker X’s family of self-driving cars are in a sense, a family. They have familial roots. This could result in them collaborating in a manner that they would not normally do with self-driving cars outside of their family. The auto maker Y’s family is a different family. The differences between the family of X and family of Y can appear in the act of driving and performing driving tasks on the roadways.

The communications aspects can either further reinforce the family, or it could potentially allow other families to become more family-like with each other.

When the right-turning self-driving car sent out the V2V about wanting to make the right turn, it could have also asked the self-driving car behind it to please act as a traffic buffer. In that case, the other self-driving car might go against its “normal” or default tendencies and opt to serve as a traffic buffer. Of course, there is no requirement that the other self-driving car has to comply with the request. It might turn down the request and basically say, hey buddy, you are on your own there, good luck making the right turn.

The nature of how V2V is going to work is still being figured out. It is relatively easy to agree to the protocols about what kinds of messages will be sent out. Trying to also agree to the meaning of the messages and whether or not other self-driving cars and their AI need to abide by things like requests, well, it’s a lot harder to settle those aspects.

Conclusion

I know that some readers might misinterpret my remarks and inappropriately commingle together this notion of kinship and bloodline as though I am suggesting that the AI of self-driving cars is going to be sentient.

It would almost be too easy to take that tack. If we said that AI will become sentient and it was therefore presumably like humans, I would guess that we’d all more easily accept the idea that they would therefore have kinship with each other.

I am not at all using the sentient get-out-of-jail free card, which is often used when you cannot otherwise find yourself out of an AI related pickle.

I believe it is a pretty evident open-and-shut case that the AI of one self-driving car maker is going to likely be a better “fit” in terms of driving behaviors with the self-driving cars of its own ilk. I suppose an AI developer could go out of their way to make me wrong on that aspect, and purposely try to make their AI self-driving cars combative with their own kind. Good luck on that. I doubt it’s going to get you many brownie points.

We need to consider the ramifications of having different styles of self-driving car driving, meaning that the AI’s will differ across the auto maker or tech firm making the AI. Though they will all presumably be driving in a legal manner, there is a lot more to driving than just driving by the law. The latitude and how you act and react to other drivers, and what you do in various driving situations are based on driving behaviors.

During the time period when we will have AI self-driving cars mixing with human driven cars, the ante is increased because now you’ll be faced with AI with other AI, or is it AI versus other AI, along with AI with humans and their driving, or is it AI versus humans and their driving. We might get some amount of road rage by humans that don’t like the driving behaviors of some of the AI’s. If you are intent on acting out toward an AI self-driving car, you’d better think twice, perhaps the AI developer planted a road rage reaction routine and you won’t like what happens once you activate it.

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

For my Forbes.com blog, see: https://forbes.com/sites/lanceeliot/

For my Medium blog, see: https://medium.com/@lance.eliot

Copyright © 2019 Dr. Lance B. Eliot

Written by

Dr. Lance B. Eliot is a renowned global expert on AI, Stanford Fellow at Stanford University, was a professor at USC, headed an AI Lab, top exec at a major VC.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store