Why AI?
There are a number of points in human history that are widely regarded as world-changing inflection points of innovation. The advent of written language, modern computing, nuclear technology, and the internet are just a few examples. But the agricultural and industrial revolutions are arguably the greatest of all. Both of these didn’t just accelerate progress, they singlehandedly set new and massively more exponential rates of progress and change. And I’m here to tell you that AI will make the combined effects of all the aforementioned pale in comparison.
I’m not the first to make such a grandiose forecast, and you’re not the first to be smartly skeptical of it (if skepticism is indeed what you’re exercising right now). However, there is a strong rationale behind this prediction, and it’s one that has been publicly endorsed by many intelligent and powerful people. Elon Musk has stated that AI poses a serious “existential risk” to humanity; Demis Hassabis (CEO of Google subsidiary DeepMind) has stated that DeepMind’s goal is to create AGI (Artificial General Intelligence) and through that alone quite simply “solve everything”, which is a goal shared by direct competitor OpenAI; and both Elon Musk and Bill Gates (among many others) have endorsed the book Superintelligence, which is in essence a detailed dissertation on how we as a society might influence the rise of a superintelligent AI to go in our favor.
It’s established, then, that some smart people feel AI is worth taking seriously — but why? This is a difficult topic to broach, but I will attempt to scratch the surface of it before moving on to explain my personal reasons for becoming directly involved in AI research and development.
First and foremost, I feel it is important to have self-awareness of the fact that our perspectives are skewed due to living in a tiny slice of human history. As a thought experiment, imagine yourself as an intelligent but otherwise ordinary person living in 1700s England. Now imagine that I tell you that right around the corner lies a great technological revolution in which fantastic machinations will assume much of the labor and handiwork currently done by men. GDP, which has increased only around 25% over the last 300 years, will increase by several thousand percent in the next 300! You’d be perfectly rational to think I was crazy, and yet, this is of course exactly what actually happened.
It’s easy to be drawn in by that dramatic spike at the dawn of the industrial revolution, but I encourage you instead to focus on the relative flatline that comes before it — how many human lifetimes could we fit in that space? Even the dramatic spike that is the industrial revolution is several times greater than the span of a lifetime. And if this is the way things have been for your entire life, why would you expect them to suddenly be so different?
It’s a misleading byproduct of observation, you see: our time is short, and in it we collect only a small picture. As a consequence our wider perspective is inevitably flawed.
So perhaps we can suspend our disbelief, at least for the moment, that such a dramatic revolution could come again. But why should we believe that AI would be the harbinger of that revolution? Because homo sapiens went from being just another species to a world-dominating one capable of spaceflight and nuclear warfare by virtue of an edge in intelligence that is, on an absolute scale, tiny — and AI is posed to imminently dwarf this advantage.
How is this even possible? In essence, it is the simple fact that technology moves faster than evolution. The human brain took millions of years to evolve, and in under a century the field of AI research has already superseded its capabilities massively (albeit only in a few narrow domains for now). Even this is not the entire story however. I am aware of at least two ways that AI could rise to intelligence levels far beyond human ones literally overnight. What’s more, both of these are not only plausible, but arguably highly likely.
The first is what Eliezer Yudkowsky terms as a seed AI: an AI which can modify itself in order to enhance its own intelligence. How quickly the AI can improve itself will of course depend on how intelligent it is, and therein lies the crux: with each improvement, it accelerates towards the next. Thus a seed AI’s increasing intelligence would be exponential, and eventually it would exceed our own intelligence by a wider margin than that by which we exceed ants. Depending on where this curve starts and how steeply it climbs, this process could take decades. However, given that we don’t know how efficient a hypothetical seed AI’s software would be, nor the hardware that runs it, it is plausible that it could take days, hours, or even seconds.
The second possibility, which is not mutually exclusive with the first, is that an AI would be created that is massively smarter than humans from the moment of its inception. It is easy to think that technology is far removed from this point and that it would take at least the better part of a century to get there. This is possible, although it is about as likely that it could happen sometime within the next few decades.
How and why? Because no general intelligence algorithm currently exists, we don’t know what the efficiency of the first general intelligence algorithm will be. However, recall that there are powerful entities actively working on making AGI a reality (namely DeepMind and OpenAI). When the last piece of this puzzle falls into place, the resulting AI could just as easily have the computational speed of an insect or something orders of magnitude greater than the greatest human geniuses that have ever lived. On an absolute scale, the range between a village idiot and an Einstein is tiny, and so too are the odds that the first AGI would fall somewhere in that range. It is far more likely that its speed of thought would be orders of magnitude lesser or greater than our own.
What will the world look like after the rise of a superintelligent AI? Assuming that this AI has even the slightest bit of physical control over the world, it would no doubt be able to sculpt the entire world to its will — just as humans already did with just a tiny fraction of the intelligence. The answer to this question then depends almost entirely on what the will of this AI is, which is in itself a deeply nuanced question. There’s one thing here that cannot be questioned however: we have to get it right.
If by some error a less than benevolent superintelligence were created, it would singlehandedly be the end of our species. There would be no contest, no resistance force, because no such thing is possible against something so vastly superior. Our first shot is also our last.
Fortunately, we live in a highly specific window of time, one even narrower than the dramatic explosion of the industrial revolution shown earlier. There are 7.5 billion people alive today, which is estimated to be less than 10% of the number of all people that have ever lived. That 7.5 billion figure will also hopefully be a far tinier percentage of the number of all people that will ever live. It is however only this several billion, in this narrow window of time, that have the ability to sculpt the future. Very few are aware of this, and fewer yet will actually try to do something with it. In essence, this is what drives me towards AI: because if I have the opportunity to play even a tiny role in such a massive and important revolution that can transform innumerable lives to come, who am I to do anything else?
Two years ago, my mother lost her battle with cancer. On the day before she passed away, I told her that I wanted to pursue a career in AI since I thought it could have a very profound positive impact on the world. It was one of the last conversations I ever had with her. I still remember that moment vividly and treat it as a promise. For me AI is also about keeping that promise and honoring her memory.
In conclusion there are many factors that drive me towards AI, and which I hope might motivate you to do the same. For one, I didn’t even mention the fact that building such amazing systems is also incredibly empowering and fun. There’s a very important revolution just around the corner, and it is my hope that we will all play our part in it.
