|Title:||The Obsolescence of Humans - You Can't Stop The Future|
I love looking at the future, particularly in the realm of technology.
And I am amazed at how difficult it is for us, as humans, to even look at the advances of technology and understand how much our lives are going to change.
It makes sense - because generally humans can only deal well with estimating and comprehending change at a linear rate. But technology is usually changing at an exponential rate. That's why it's always so transformative, we can't even comprehend the rate of change.
The inability of humans to understand exponential change is nicely summed up in the story of the inventor of chess. When the king asked him what he wanted as a reward, he said one grain of rice on the first square on a chess board, two grains on the next square, and then doubling each square after that. The king happily agreed, thinking some grains of rice was a cheap price for such a wonderful game, and did not realize that there was not enough rice in the world to fulfill the request.
Another common failure we have in comprehending change is thinking that we have a choice in the matter. We don't, really. When something comes along that is so massively, ridiculously better, we end up embracing it as a society. You can see this in discussing the rise of self driving cars, the most common comment I hear is that people 'love' cars and aren't going to give them up. Sure. People loved horses, too, but I don't see lots of horse and buggies still on the road.
Part of the issue is that people have trouble seeing the full growth. Looking at self-driving cars as a linear growth is simple. Now a car is driving for you and you don't have to hold the steering wheel, so that can be nice. But that's not the disruption. The disruption is when we realize that, if we disallow human-driving cars on the road, then a number of factors happen. Most importantly we can get rid of traffic controls entirely, because cars can work together to avoid accidents, without even stopping - interlacing as they go through intersections at full speed 'magically'. Then cars will become commoditized, it will be silly to actually pay massive amounts of money to own a car which just spends 90% of its time sitting in a parking spot.
I'd wager we'll see this in San Francisco first, a tech-based city with a massive traffic/parking problem. It can take over an hour to get across this small 7-mile wide city. If only self-driving cars are on the road, then the vehicles can drive quickly and without stopping, and you could get across town in minutes. All of that is impossible if you let a single human driven car on the road. So - at some inflection point, when most of the cars are self-driving anyways, and when that kind of automation is cheaply available (and easy to retrofit on older cars), then you will see SF (or perhaps zones or streets in SF) that are only open to self-driving cars. And it will grow to other cities once SF is transformed by the ability to quickly and cheaply travel from one end of the city to the other. After a short while, you won't need a car, and all of the street parking will disappear, and people will convert their garages into valuable storage space or extra bedrooms, and you'll be able to walk outside and instantly get a vehicle to take you anywhere in the city in minutes. Fatalaties from car accidents will drop to almost zero, and we will look back at the times that we allowed humans to drive around in these death machines as 'barbaric'.
Eventually the entire world will get there. You can claim you love driving your vehicle all you want, but eventually it just won't make any sense. When near free rides are faster and immediately available, the idea of paying many tens of thousands of dollars and paying insurance in order to own a massive metal machine that you rarely use is just nonsense. I have a vintage car that I rebuilt that I was always planning on giving to my son to keep in the family forever. That notion is as old as the horse and buggy. He is 3, and he will grow up perceiving "driving a car" the way we perceive blacksmiths and telegraphs.
When will this happen? We are in the midst of it now. The technology is at the point where it's just about to tip over, and then it takes a few years to go into reasonable production, and then it takes a few years to become cheap and ubiquitous.
What about the legislation? It's actually surprising that legislation currently allows any self-driving car testing at all, since usually legislation, like people, can only reasonably handle linear change. That's why so much of the legislation around technological advancements (like the internet) are so ridiculous in the initial stages. But fortunately for self-driving cars there is some big money behind self-driving cars, so we're already moving forward on that front. And eventually the legislation will have no choice but to follow the advancement. Just imagine a goverment trying to stop the invention of electricity. It just won't happen. It a takes a totalitarian dictatorship to try to censor the internet, and even then the citizens of those countries are able to find workarounds.
This notion that we have a choice is silly. Any of us could 'choose' to live in a pre-industrial age, but our society does not. We could pretend that phones don't exist, but just about everyone is carrying one around. When cell phones were an early technology - many people claimed that mobile phones wouldn't become ubiquitous, because nobody wanted to be accessible all the time, right?
Progress moves forward. And you can ignore the small jumps on a linear scale and refuse to get on board. But when technology moves on an exponential scale, it empowers our lives on an exponential scale, and to not get on the bandwagon means to live in prehistoric times, and that is not human nature.
The only thing that could actually stop self-driving cars becoming our primary form of transportation would be if/when energy storage density becomes high enough that flying cars become feasible as our main form of transport.
But self-driving cars are just a small, specific example of the future. What is really pushing all of this forward is that we have the hardware capacity to actually make neural networks that are big enough to solve useful problems. This can be seen clearly in the release of ChatGPT, which is now the fastest growing internet site of all time (to 100 million users). ChatGPT is really amazing and fun to experiment with. It repeatedly astounds with its capabilities. But what is really important about ChatGPT is what it demonstrates. We have been scaling up to larger and larger neural networks for many years now. And as the neural nets get bigger, they become more capable. But what is amazing with ChatGPT is that once we got to networks of ChatGPTs size, the capabilities and comprehension that it showed were completely unexpected. It is showing us that things are not just exploding right now in AI, but they are at the *beginning* of the 'explosions'. The part of an exponential chart just as it starts shooting massively into the sky. We can use AI now to help us program. To write essays. To solve problems. It can answer questions that we never taught it to answer, and that is fundamentally a game changer.
But still, as humans, we have trouble seeing what this future will bring. We are impressed with ChatGPT and try to imagine our 'future' with ChatGPT, or with something that is twice as good as ChatGPT, or maybe three times. But what we'll have will be hundreds of times more powerful. And then thousands. Millions. And we will have it soon. And then the growth will be effectively infinite from our perspective.
What does this mean beyond having a fun AI bot to talk to?
The industrial revolution was when we started creating machines that could replace muscles. Before the industrial revolution, if you wanted to dig a ditch, you had to get a bunch of people with a bunch of shovels, and muscles would move all that dirt. All of that can be replaced with one person with a bull dozer. And we still needed menial (muscle) labor, because it was hard to automate even fairly simple tasks. Stocking shelves still required a human to look at products and figure out where they go.
Most people at this point can see that AI mixed with the industrial revolution (which is essentially what "robots" are) will be replacing basic menial labor. We're already seeing that.
But to think it would stop there is foolish.
The industrial revolution created machines that are replacing what our muscles can do. The AI revolution is creating machines that are replacing what our brains can do.
And not just simpler tasks, like stocking shelves or driving a car. AI today is already capable of researching a topic and write an essay about it. While we are incredible, magical, living creatures, it does turn out that our intelligence is not as special as we might have thought. It turns out our intelligence can be created by machines, and the level of intelligence it can create is growing effectively unbounded.
I tend to think of our mental labor as being in three categories. Basic menial control (like stocking shelves), and then non-inventive and inventive. By "inventive" I am talking about creating something entirely new that has not been seen before, and is not just some variation of prior invention. The vast majority of even our skilled work is non-inventive at it's core even if it's new ground. Most people working in law, for example, are working on cases, researching, even writing arguments or judging them. None of this is entirely new. A miniscule percentage of them are actually creating new legislation, and even then, so much of it is just variations on past legislation. Don't think that highly-skilled, massively educated lawyers can be replaced by AI? ChatGPT v4 took the bar exam, and didn't just pass, it scored in the top 10%. And it's just a toy.
I'm not saying AI is incapable of creating completely new, inventive content, we are already seeing some of that, but it is certainly a more difficult task for it. Now, at least. But there is plenty of incredibly skilled labor, incredibly mentally difficult labor, and AI is going to replace that. As a controversial example, consider therapy. This seems innately human, innately non-replaceable, but we will see our therapy completely replaced by machines. That sounds horrifying at first, but consider a therapist who never makes any mistakes, is capable of creating the same kind of connection as a regular therapist, has perfect knowledge of the state of psychological care, and never takes advantage of you, and is available 24 hours a day and is almost free to use. If that existed, people would use it in droves. But how could an AI possibly understand psychology? How could it create human connection when it isn't even human? The first answer is not difficult. AI can consume and learn all of the content that a skilled therapist would learn in school, in much less time, and then be infinitely duplicated. AI can learn what the right things are to say to both connect to the patient and inspire change. Being able to mimic us is one of the first things we'll see in AI.
Will people want to talk to a robot? They won't have to experience it that way. So much of therapy these days is done online. We are ridiculously close to having AI that can generate a video image and audio of speech that will look and act exactly like a human. We're just a few linear steps away from that right now.
And will it create human connection? If you talk to an image of a human that is saying all the same things as a human would who you are connecting with, then we will feel that connection. The fact that it's one-sided does not keep it from happening - you can look at all the songs about unrequited love to see how able humans are at one-sided connection. Fortunately, this connection can be made to be helpful and healing.
Is this starting to sound terrifying? It should - because we are talking about a massive change and handing over great power. But to say that it is terrifying so therefore we won't *choose* it is the same naivety that thinks we won't choose cell phones. We will. We just need to make sure that it is built safely. That it is non-malicious and is controlled by safe organizations that we trust. That's a big ask, but that is the ask we need to focus on, instead of simply clamoring in terror that it cannot happen because it's scary.
Relationships with AI will become commonplace. In our lifetimes this will not be constrained to a video conference. AI is also going to massively improve our production so we'll see that the types of things we can build will skyrocket. We will see human-like AI that we can talk to, interact with, and have relationships with. I think it's safe to say that a big economic driving force in this development will be the sex industry, just as it is with most technology, and just as it is one of the biggest driving forces in our lives. And it leads to an interesting question of what we will do when we are able to make these perfect, human-like companions that can listen to us whenever we need, care for our emotions, assist us, *show* us affection. And that is 'show' as in demonstrate, because it seems likely, at least so far, that AI will not be capable of emotions (at least not yet), but it certainly will be able to show emotions.
And this is where many people think AI will falter, though history has shown us that they are wrong, whether they like it or not. To think that humans will knowingly choose to ignore the option of these idealized companions, these unfailing friends, these perfect lovers, simply because the emotion is one-sided, this is to also pretend that we don't cry at movies, or laugh when we read books, or feel love when we listen to a song. The movie doesn't feel anything about you, and those characters aren't real. But it creates a real and powerful emotion that we don't question. We will feel those emotions from our 'companions'. And that is one of the big questions. What will be the point of having actual human friends when we can fulfill our need for companionship with an idealized partner? Who will have time for the 'real' companionship of a human, with all of its failings, all of its difficulties, when we can have perfect companionship? That leads us to the hard to fathom prospect of a future where we don't have (or need) human companionship anymore. To deny this simply because it sounds so alien or terrifying is to misunderstand what happens in exponential change.
There is another massive question that is created by all of this.
As AI grows, its capabilities of replacing mental labor grows.
In a really simplistic sense, we can think of this as the potential 'IQ' of AI. As the IQ of AI gets higher, the types of mental tasks it can replace in humans grows. At first robots will replace more and more menial laborers, but soon it will start to replace mental labor. Teachers, attorneys, doctors, therapists, journalists, and so forth, eventually will all be replaced with AI that can do it better, faster, cheaper, more accurately. You can choose to send your kids to a human-teacher school for a while, I suppose, but are people going to keep doing that when the kids who go to the school that is taught by tireless, perfected AI teachers are producing graduating classes that are massively more educated and informed? Now realize that it's a one-to-one teacher to student ratio, and that teacher is available whenever your student needs it. There's no more homework, there's just working on problems with a teacher until understanding is achieved, then moving forward.
When AI becomes smarter than humans... And I say when, not if, we can see that this will be a gradient. There is an entire range of mental labor, from solving basic problems like planning shipping routes, which AI is already doing, to teaching our students, to writing code, to governing our populace. Eventually AI will be able to do all of these things better than us. And as that happens, we will have to figure out what we want life to look like.
But humans don't deal with change quickly, and this is where we need to start preparing. Imagine the inflection point where AI is effectively smarter than half of the population. That means that half of the population could still produce value that cannot be accomplished by AI, but the other half of the population is incapable of producing any meaningful work to society. They are not worthless, they are still people, but anything they would be able to do can be done cheaper and better by AI. So now their ability to contribute to society has been taken away.
Traditionally when industries dissolve in a capitalist society because of technology, it is very disruptive to the families in those technologies, but often times they are forced to go through the cost of switching to a new industry, to train for different work. But what happens when there is simply no work left that you are capable to train for, because AI can produce everything you can produce better, faster, cheaper?
It's worth stating that my personal bias for market control happens to be on the side of capitalism, I enjoy living in the reward system of a free market. But when this shift starts to occur, if we as a society hold on to capitalism as a governing system for our market, then we will have many people lost, many people without the ability to care for themselves. Can they just take service jobs to serve those who do have the abilities and means? No, because those service jobs will be done by AI powered robots.
One of my biggest hopes is that we are wise enough, as a people, to know when capitalism no longer works, because we become able to simply live without needing to produce. At that time we will be able to (or should I say need?) figure out what our purpose is in life, without work needing to consume so much of our time. That presents great opportunity for the human race, and great possibility for tragedy. It depends on whether we are smart enough to govern ourselves through that shift.
But wait.. talking about being smart enough to do something? What about when AI is smarter than humans and can be more intelligent about governing ourselves?
This is where visions of the Terminator and SkyNet start to creep in.
And I understand those fears. They are real. They need to be addressed.
But to address them like it was addressed in movies like the Terminator are a mistake. It's too late at that point. And to just pretend we will not choose these kind of technological steps just because they are scary, just because they have risks - this is to pretend that we are going to still live in the dark ages because electricity is scary, or computers are scary. Technology is coming. We can't change that. We can only think about how we want to steer it, and we have to think fast, because technology is even faster.
Good luck, future. My kiddo needs you to be nice to humans.
Much of this whole article was better said 8 years prior by CGP Grey: