Speculations on the impending era of artificial intelligence

Robot_AI

The relationship between man and machine has been a very fruitful one for our species mainly because it has empowered us to transcend many of our limitations. I remember coming across a Scientific American study a while ago that nicely illustrated how well humans have been empowered by our technology. The study measured the efficiency of locomotion for various animal species on the planet. In that study, the condor was found to be most efficient… using the least amount of energy to travel one kilometer. Comparatively, we humans came in with a rather abysmal showing somewhere way down the list… not exactly the most flattering result for the supposed crown of creation. However, when some brilliant person at Scientific American had the insight to test the efficiency of locomotion for humans on bicycles, we blew the condor away, climbing to the very top of the “efficiency of locomotion” charts. Thinking deeper, we start to realize that in an analogous way, our technology has served as a metaphorical bicycle for our limited physical and mental abilities. For example, the advent of the internal combustion engine and the automobile has reduced what would have been a multiple month long journey 100 years ago to one of just mere hours as I type this. Also, the next time you open up a spreadsheet on your computer to do some quick calculations at work, think about how much longer that would have taken if you had to use a slide rule or an abacus. It is likely that the same task would have required a ton more effort and probably taken 10 to 100 times longer if you had to do it in say the 1930’s when computers didn’t exist. These kinds of miraculous time savings have allowed us to be more productive while simultaneously conserving more mental cycles to devote to things at the higher echelons of mental function.

As far as we’ve been concerned over the past century or so, our machines were supposed to just sit quietly in the proverbial corner and do what we told them to. When we punch a series of mathematical operations into a calculator, we expect the answer to that exact calculation… not what the calculator “thought” was best for us at the time. When we command our phones or computers to play a particular movie or TV show episode on Netflix, we expect to watch that exact movie or episode… not what Netflix “thought” was best suited to that particular occasion. For the most part, we expect our machines to work without thinking about it… whatever “it” is. So far, we’ve thrived very well on this “master – slave” relationship between humans and machines, becoming much more powerful than any of us could have ever fathomed in the process.

Recent advancements in the field of artificial intelligence (AI) are slowly but steadily altering the nature of the interaction between man and machine. While there remain many instances of interactions between man and machine in which the human is still the unquestioned master to the machine slave, there are many emerging technologies that are starting to buck this trend. An everyday example of this is when you use the Google maps application on your smartphone to plan a trip to a place you are travelling to for the first time. All you have to do is type in the destination address and hit the return key. Once you do this, the maps application will quickly determine the fastest route to your desired end destination… taking traffic congestion and any reported accidents into consideration. If you have one of those cars that connects with your phone, you can now just sit back and follow the turn by turn directions prescribed by the humanoid voice coming from your phone through your car’s speakers. With a bit of thought here, we should all be able to see how dramatically the roles have been reversed. The smartphone is now the intelligent one that thinks ahead to plan the trip and is actually able to intelligently alter your directions on the fly as faster routes to your end destination emerge. In this scenario, you and I as humans are now relegated to being the “dumb” part of this relationship… all we have to do is sit there and try not to accidentally crash into things along the way.

Alright so artificial intelligence has become better than we are at selecting driving routes, but we’re still superior to artificial intelligence in the realm of learning… right? Surely no computer program on the face of this earth could possibly defeat a chess grandmaster at the game of chess for example… right? Well, actually, no… wrong. Artificial intelligence has matured to the point where some of the best ones can consistently defeat the best humans at the game of chess which is one of the most mentally complex and challenging games ever conceived. The thing about chess (or any other game if you play it at a high enough level) is that the best players learn very quickly on the fly and can often see the game several moves ahead. In extreme cases, the victim of a chess grandmaster may not realize they’re already defeated until 3 moves into the future even though the chess grandmaster would have known all along. Taking all this into account, it is pretty impressive that a software package can somehow pull off defeating a chess grandmaster at his or her own game.

One of the major reasons why artificial intelligence has gotten so smart at specific tasks is due to a process popularly known as machine learning. Machine learning is very analogous to the way we humans learn. Remember how many times you fell off your bike when you were first learning to ride it? Yeah I know… I fell off a bunch of times too. Neither of us quit though… we kept going and correcting our errors after each fall until we mastered it. In an analogous way, machine learning is an iterative process in which a machine can “learn” to be good at something by trying to do it a bajillion times, sucking at it to begin with, but constantly learning from its mistakes until it attains mastery. Like you can probably imagine, the more practice runs the machine gets at a certain task, the more adept it becomes at said task. In the same way, a chess centric artificial intelligence software package (chess AI) is probably pretty crappy at the game of chess to start with. However, it can play hundreds of millions of games on a continuous loop using the immense computational power at its disposal, learning from its mistakes and cataloging the appropriate strategic responses to all possible situations on the chessboard as it goes along. I think you can see that we humans cannot match this intense level of “practice” that the chess AI can sustain. While it might take a human 10 years to rack up 10,000 hours of practice at chess, our hypothetical chess AI can theoretically do it in a little over a year.

This ability to learn at a supernaturally fast rate is one of the reasons why artificial intelligence holds so much promise for facilitating a brighter future for humanity. To drive this point home, let us Imagine an artificially intelligent entity whose sole purpose is to completely master every last nook and cranny of the human body and all its functional organ systems. We’ll call this imaginary AI “human body AI”. Human body AI would spend every moment scouring all medical databases and “reading” all peer reviewed journals in the medical field, cataloging and internalizing all the important discoveries as they happen. In a few years, our imaginary AI may have amassed enough information to be able to build the most accurate simulated version of the human body upon which it can perform virtual experiments that would closely mirror reality. As a result, human body AI would probably have a lot more knowledge about the human body at its “electronic fingertips” than any human ever could. With that much power, who knows what could result? Perhaps human body AI could use the superintelligence at its disposal to conceive of a cure for pancreatic cancer or AIDS… solutions to two of the biggest medical puzzles that have eluded us partly because of the limits to memory and recollective power that we humans are constrained by. If ever something that revolutionary were to happen, it could permanently change our world for the better. Think about how many lives we could save. Think about how much suffering we could curtail or prevent altogether. The potential for good here is absolutely enormous.

Although we can see much beneficial potential with AI, there is also an equal measure of catastrophic potential associated with it. If we as a species give birth to something as potentially powerful as a polymathic artificial intelligence without being able to sufficiently contain and control it, the consequences could be apocalyptic. The current forms of AI are relatively harmless because they usually can only do one thing very very well. Miss “Siri Google Maps” is only good at telling you how to get from point A to point B, but she can’t control your electricity or beat you at a game of chess (if you’re a grandmaster). The passage of time however could give rise to polymathic forms of AI which in plain english just means AI that is excellent at everything. In the minds of many, giving rise to a polymathic artificial intelligence without the ability to sufficiently control it could result in one of two outcomes. In the first somewhat benign scenario , this polymathic AI would see us in the same way we see the house cat or our pet dog: something that needs to be fed, cared for, caged, and trained to behave in a certain way. In the second more sinister scenario, the polymathic AI could see us humans as pathetic relics of a evolution gone awry, and calmly go about the business of architecting our destruction. Obviously, the second more sinister scenario is the one that mankind must avoid at all cost. It is however not clear that we as a species are taking the more sinister outcome seriously enough to put measures in place to prevent it. Perhaps our levels of intelligence prevent us from being able to clearly peer into the future and preempt this potential disaster… only time will tell.

During my first year of graduate school at Stanford University, one of my most revered professors once said: technology will continue to progress whether you like it or not… it is best to be prepared for it rather than waste your time complaining about it. I have found those words to be borderline prophetic in the 10 years that have elapsed since he uttered them. It would behoove all of us to heed his wisdom and apply it to our lives as we stare down the impending AI revolution. As self driving cars become a more prominent part of our everyday lives, truckers and taxi cab drives will lose their jobs to a combination of a fleet of self driving cars and cleverly designed smartphone applications. Even software engineers who are seen as perhaps the most valuable skilled workers in the world today may also be eventually replaced by artificial intelligence. It isn’t outside the realm of possibility for computers to very well be able to someday effectively program themselves. The only jobs that may yet remain safely in the hands of humans for the foreseeable future are those that require a generous injection of emotion and intuition. Artistry, Musicianship, Creative Writing, Poetry, Singing, Dancing etc. Emotions are still too beautiful and complex for any machine to effectively replicate… for now anyways.

Although my colleagues and I are very excitedly looking forward to the artificial intelligence revolution, we do so with a heavy dose of trepidation… perhaps even fear. The last thing we unleashed with this much potential for destruction in the annals of history was nuclear weaponry. As we all know, that didn’t go so well but at least we had the power to dial it back. Developing polymathic artificial intelligence reminds me of the insane dude in the stereotypical horror movie who wants to summon the demon because he thinks he can control it for his benefit. Nine times out of ten, the demon ends up controlling him. Let us all hope that we approach this potentially seismic shift in the course of human history with the humility and prudence it deserves. Our lives may very well depend on it. From all of us here at chubaoyolu.org, please take care of yourselves and each other.
 
 
Without Wax
Oyolu B.C. Ph.D.
chubaoyolu.org
Visit our ETSY store!

2 Comments »

  1. Well written and insightful. The effect of AI on medicine alone can potentially alter the course of human history in our favor. Or we could face extinction in a war with an AI that wishes to eliminate us. Fascinating. Thanks for writing.

    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s