Yikes, Lack of Geniuses Holding Back AI Advances, Driverless Cars Say Me Too
Dr. Lance B. Eliot, AI Insider
Is there a genius shortage that is impeding the progress of AI?
This is a pointed question that keeps coming up in the hallways of AI conferences and that people are whispering about. Sure, there has been some impressive efforts of newer AI systems that suggest we are making solid progress in AI, but it’s not particularly breakthrough-like improvements that have rocketed AI ahead and overcome some of the as-yet-solved thorny problems in AI.
Some people react by saying it is a blatantly stupid question. What does being a genius have to do with progress in AI? Do we need to have an Einstein of AI, or a Darwin of AI, or a Leonardo da Vinci of AI, in order to push further ahead on AI?
In essence, you might make the argument that by-and-large the progress in most fields of endeavor has been undertaken by “less than geniuses” that did the hard work and painstaking efforts to make progress. Lots of really smart people can perhaps do the work of those unicorn geniuses. Historians would likely indicate that those of a genius nature are far and few between, and you’d be unlikely to pin substantive progress in endeavors primarily due to those geniuses alone.
What Is Genius Anyway
Some say that genius is in the eye of the beholder.
You might see someone do something and remark that the person is a genius, yet others might smirk that the person was not at all a genius and you were fooled or misled or misunderstood and assumed the person was a genius. You might be tempted to use IQ as a measure of genius and suggest that when you have a certain high number of an IQ that you are ergo a genius.
At the AI lab you work at, right now, look around, and there might be a genius to your left or your right, on their way toward AI genius breakthroughs or perhaps will do so in a few years (or, if you prefer, look in the mirror!). These budding geniuses might be like the moth that will someday emerge as a butterfly, allowing their inner genius to make its way out and astound the world of AI by solving seemingly insolvable problems.
There is also the matter of genius as a sustained trait versus a transitory or eureka kind of flash of brilliance.
Hindsight and the writing of history can at times bolster the case for someone being considered a genius. We might then fall into the trap of assuming that the genius flavor was the genesis for an amazing insight that others never had. In fact, many times the alleged insight was one that others also had at the time, and for a variety of confluences it turns out that the one person now having fame as a genius gets the glory, though many others were doing similar work at the time.
If you are going to claim that there is a genius shortage, it implies that there is some magical number or desired threshold of geniuses that we are hopeful of attaining.
Economic Supply and Demand of Geniuses
How many geniuses do we need in AI?
You could ask the same question of any other field of inquiry. How many geniuses does physics need to make abundant progress in physics? How many geniuses does chemistry need? And so on?
A recent research paper weighs into the genius shortage debate by trying to model the level of societal genius in an economic manner.
Research by Seth Benzell and Erik Brynjolfsson at MIT provides an interesting look at the so-called G factor, an economic parameter associated with genius. Their study entitled “Digital Abundance and Scarce Genius: Implications for Wages, Interest Rates, and Growth” examines genius as a limiting factor in economic growth. They point out that though the advent of our digital world has allowed labor and capital to become more abundant, we are still limited due to the inelastically supplied complement of human genius. For their paper, see: http://ide.mit.edu/sites/default/files/publications/Digital%20Abundance%20and%20Scarce%20Genius%20for%20shortened%20abstract.pdf
By considering the matter overall as an economic one, it is a modeling exercise of having an out-of-balance of supply and demand, in geniuses, an otherwise scarce commodity.
We have a demand for more geniuses, seemingly, and our supply is too low. Its time to cultivate those latent geniuses. Find them, spur them on.
Making Geniuses Via AI Is Another Path
There’s another path that some AI developers are hoping for.
Maybe we can craft AI that has genius, therefore we won’t necessarily need as many genius people or at least not a lot more people to have genius in order to meet the lack of supply of genius. Use digital technology to make geniuses, either by the AI itself being a genius, or perhaps if that cannot be readily done then at least be an aid to boost humans into becoming geniuses.
On the matter of the potential for AI systems geniuses, worried humans are concerned that this might take us down the path of having an AI singularity or a super-intelligence, and for which we as humans might then become their slaves. Or, the AI might decide to wipe us out entirely. Is this merely conspiracy kind of talk? Or, is there merit to the dangers? Something worthy of further debate.
AI Self-Driving Cars and the Genius Shortage Topic
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. One looming question for the auto makers and tech firms is whether or not there is a need for geniuses to make the advent of true AI self-driving cars become a reality.
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.
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.
Genius Shortage and Impact on AI Problems
Returning to the topic of a genius shortage in AI, let’s consider the nature of AI problems that need to be solved and whether we’ll need genius thinkers to solve those problems.
AI of today lacks common-sense reasoning. This severely limits the ways in which we might make use of AI.
Some also liken this to Artificial General Intelligence (AGI), namely having a type of AI that applies across domains and is not focused or fixated on a particular domain. The AI efforts to-date are primarily narrow in their scope.
For driving a car, there is an ongoing debate about whether or not the AI needs to have AGI. Human drivers do have AGI, therefore if the AI for a self-driving car is trying to drive like a human, presumably the AI needs to have AGI. Others claim that the driving task is narrow, and therefore there isn’t a need to have AGI for the self-driving car driving task.
What about common-sense reasoning? Humans use common sense reasoning when they drive a car. Does an AI self-driving car need to have common-sense reasoning, similar to the kind of common sense that humans have? Some say that the AI does not need overall common-sense and that it can be programmed sufficiently to have similar qualities.
If we had more geniuses in AI, would we by now have solved the AGI problem?
If we had more geniuses in AI, would we by now have solved the common-sense reasoning problem?
I don’t know how we can answer the question, since of course if you define genius as someone that would have solved those AI problems, the answer is that yes, those problems would be solved by now if we had them around.
We also don’t know that there is a magic formula that a genius would discover to then solve those AI problems.
More Open Problems in AI That Genius Can Tackle
Another AI open problem involves the topic of learning. Today’s Machine Learning and Deep Learning is actually shallow when compared to human learning. How can we get Machine Learning or Deep Learning that can do one-shot learning, whereby after experiencing only one or a few examples the AI is able to generalize and learn about a topic or matter? It’s a tough problem to solve.
There is the open problem of getting AI to learn-to-learn. We cannot keep setting up AI systems that we, the humans, have done the construction around what the AI will learn.
Some even suggest that we need to have the AI begin as a kind of child-AI, and let it grow over time, similar to how humans start as babies and become children and become adults. Perhaps that’s the only way toward getting AI that is more robust.
Object recognition is another open problem in AI.
Today’s AI systems that do object recognition are not doing the same kind of “recognition” that humans do.
Autonomous navigation is another open problem in AI, including advances desired in SLAM (Simultaneous Location and Mapping).
Theory of Mind is another interesting and important AI problem. Humans that interact with other humans will tend toward having a Theory of Mind about the other person, being able to guess what the other person might be thinking about, or what the person might do in a given situation. The AI systems of today do not have much if any of a Theory of Mind embodiment.
In the case of AI self-driving cars, when human drivers are driving a car, they are usually anticipating the actions of other drivers. You watch the car ahead of you and can guess that based on their driving behavior, they are a timid driver and therefore you can anticipate other potential driving moves the person will make. Few AI self-driving car efforts are seeking as yet to embody this kind of capability.
Across the Spectrum of AI Self-Driving Car Elements
I’m not going to enumerate all of the open problems in AI, but I’ve provided some of the more notable ones and hopefully provided you with an indication of where there might be value in adding a “genius” to try and solve those problems.
The sensors of an AI self-driving car are the key to sensing what is around the self-driving car. Maybe there are new kinds of sensors that nobody has yet even invented or considered, for which a “genius” might come out of the woodwork and create.
We might be lucky to have a “genius” that can vastly improve sensor fusion. The capability of cohesively bringing together the sensory data and make sense of it, well, it’s a tough problem. Similarly, the use of virtual world models could use a “genius” to make those more powerful and capable. The same can be said of the AI action planning portion of an AI self-driving car, and likewise for the car controls commands issuance.
There is also the need for AI self-driving cars to be self-aware. A human driver knows that they are driving a car. The human presumably keeps tabs on themselves, realizing when they are getting sleepy or impaired.
Do we have a shortage of geniuses in AI? Besides the aspect that you could presumably say the same thing about nearly all other areas of study, let’s just say that if we had more geniuses it might be helpful.
Notice that I say that it might be helpful, rather than categorically saying it would absolutely be helpful. We don’t know that the geniuses would necessarily be ones that would help us make progress. Suppose there are geniuses that are devious and opt to take us down a bad path? Or, maybe there are geniuses that are trying to do their best, and yet waste our attention on something that won’t payoff. Who knows?
If you are a genius, please jump in and help out on solving these thorny AI problems.
If you are not a genius, maybe you can become one, so please make a go of it.
If you are trying to help someone else to become a genius, try not to go too far since making a genius is not a sure thing.
If you are an AI developer seeking to craft AI-genius, good luck to you and aim to ensure it won’t wipe out humanity.
There is a famous quote by philosopher Arthur Schopenhauer, “Talent hits the target no one else can hit; genius hits the target that no one else can see” (in his book, “The World as Will and Representation”), which is worth ruminating on.
Maybe I’ve not even listed the solutions that a genius will come up with, presumably being able to see problems and solutions that the rest of us aren’t even yet able to discern. Go for it, geniuses, let’s see what you can do.
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/
Copyright 2019 Dr. Lance Eliot