The Age of the Lucky: AI & the End of Neutral Probability
You are about to become one of the luckiest humans to have ever lived.
The sort of luck that wins lotteries, or leads to your “meet cute” when you spill your drink on the woman of your dreams. The serendipitous kind of luck from movies like Serendipity, the kind of luck that’s beyond our control and can be the difference between success and failure. People will literally get struck by lightning less and hit jackpots more.
When I say you’re about to become one of the luckiest humans to have ever lived, I mean it literally — everything you think of as luck is about to change.
And I think there’s a realistic chance that this “luck” will solve world hunger, homelessness, failed marriages, and possibly world conflict. For all the discussion about AI, I feel like we’re overlooking the most important impact it will have on everyday people like you and me.
I propose that the biggest impact on your and my life will be that you never again show up at the coffee shop to discover they’re out of your favorite pastry.
Oh, and like, 90% of the world resource problems will sort of just magically go away for about 50 years.
The Age of the Lucky
The movie Her, about a man who falls in love with his AI, is one of the few movies about a super advanced AI that isn’t evil.
It’s also how I think many people imagine “good AI” will look like. Essentially the smartest person you’ve ever met doing nothing but trying to make your life better, 24-hours a day. But that vision doesn’t go far enough in describing the impact on our lives — it’s too easy to imagine.
Imagine instead that the AI from Her does exist, as it does at the beginning of the movie. You fall asleep at 11:00 pm, and it immediately organizes and optimizes your following day. Arranges meetings for you, responds to some emails for you, removes a virus from your computer, whatever. It does all of this in the first 60 seconds, and now it’s 11:01 pm. What does your AI do for the rest of the night?
The simple answer is it begins running through probabilities. It starts “pre-thinking” things. You have a meeting tomorrow morning at 8:00 am? What is the probability of rain tomorrow? Will you need an umbrella? How far away is the meeting location? Again, this level is intuitive and easy to imagine.
AI will be very good at this pre-thinking…far beyond reasonable human expectation. By 11:02 pm, it’s figured out all those obvious questions. And then it keeps going. It’s likely to rain early tomorrow, and you’ll likely decide to drive. When was the last time you had your car engine looked at? You haven’t changed your oil in…omg, the entire life of the car. Probability of failure for the car is calculated at 0.03%, which is high enough for the AI to worry about. If it does break down, what route is nearest a repair shop? What are the contingencies?
By 11:05, your AI will have “dreamed” about layers and layers of probability problems that we’d consider ridiculously minute. If you like banana cake from the café you’re meeting at, your AI may calculate the probability that they’ll have banana cake at that café each morning — which is probably likely. After all, they’ve had banana cake most of the time you go there. It’s probably better than 95% chance they’ll have banana cake. Almost guaranteed.
But for the AI with an entire night of time on its hands, it’s far too much uncertainty. So it communicates with the AI of the baker at the café, and they allocate the number of banana cakes appropriately. Now there’s a 99.6% chance that there’s banana cake — and you’ll never even know about it.
It’ll just be good luck to you.
Every element of our lives that has unattended random probability will just become more likely to fall out your direction.
From your perspective, the coffee shop just magically always has banana cake, and you can’t remember the last time that they were out. From the café’s perspective, managing their supply and demand becomes easier. And when your car breaks down on the way to work one day — because you’ve been putting off that oil change you’ve been meaning to do forever — you magically break down near a repair shop that happens to be right next to your favorite café, so you can wait in comfort while it’s fixed. There you meet someone who offers you your dream job, or a cute girl into the same show on Netflix that you are, or…
You can go to ridiculous places with this. To levels that sound so beyond realistic that it’s stupid, like going to lunch and having someone hand you a job offer doing exactly what you love for 20% more pay than your current job…out of the blue. No application or job interview needed. It almost sounds idiotic.
Yet, what’s amazing to me is that no matter how far you run it, I think it’s likely that we’re not capable of exceeding the fantastically-far reaching realistic reasonable. When every person is carrying a general AI that is optimizing life for each of us to that degree, every second, and communicating to each other about our mutual goals, I’m not sure it’s possible for us to predict how much ‘randomness’ will disappear from our lives.
We’ll just start being magically, magically lucky.
How This Will Solve… Pretty Much Everything. For a While. Let’s Start With Hunger.
Humans have been successful because of our ability to overcome inefficiency — mostly by overproducing. Food. Money. Housing. Energy. Everything that we create has to overcome a default amount of inefficiency. To feed 10,000 people, you need to produce enough food to cover their needs, plus enough to cover all the waste — estimated at somewhere between 20% and 40% of the food we produce each year.
Since the beginning of human civilization, we’ve survived because we’ve outproduced the inefficiencies in everything we do. We’ve engineered overproduction and tolerance for error as a part of course. And we’ve engineered the most tolerance into systems that have to interact with human beings. We people are freakin’ crazy, and absorbing our behavior introduces tremendous potential for error — and so we’ve built great redundancy into everything we touch.
In a simplified way, how much local hunger could be solved if we had perfect information about who within a 1 block radius needed half a sandwich at the exact moment I decide I’m not going to eat the last half of the sandwich I ordered for lunch?
From a global perspective, roughly 9% of the human population is considered food insecure. And while the complexity of production, distribution, and regulation make this a daunting problem to solve in a centralized way, AI isn’t centralized — it’s free to be driven as a traditional individual actor. How much improvement do you need across all of those sectors before the excess production and efficiency has a significant impact on that 9%?
The power of this is that it does not require the coordinated effort of the supply chain to see the improvement. In the way that self-interest can incentivize people to solve problems, an AI interested in each actor’s self-interest will solve problems in the same way. Dysfunctional governments will become more functional as critical information begins connecting through unexpected pathways that were previously invisible. People trying to solve problems will just become better at doing so, by nature of their AI and its communication to others trying to be more successful, as well.
As incredible as it seems, I personally believe that it’s difficult to overestimate the impact of hundreds of millions of individual AIs, connected globally, minimizing the probability problems in our lives, and maximizing serendipity.
Almost without effort, our current production ability will just start being better at covering our needs. Everything. Simply having AI that’s reviewing, communicating about, and optimizing for all the aspects that could impact our individual success will by default force AI to apply optimization to parts of our lives that we’d never have set out to optimize on our own.
By that time, it’s about 11:10 pm. 50 minutes to midnight.
The Return of the Single Income Household
In previous technical advances, the improvement in efficiency has had an impact on daily life through the lowering of cost of goods and services. The industrial revolution and mass production made it cheaper to make — and therefore sell — stuff. The standard of living for the average person improved as a result, faster than the improved efficiency has degraded people’s ability to earn.
I think — and hope — that we’re about to see this happen again at a scale that’s so unprecedented that it’s hard to imagine.
The difference between this upcoming revolution and ones in the past is that previous revolutions have empowered future production, and the luck of the AI will empower existing production. When I harness electricity for the first time, it empowers my future tools to use electricity. Before electricity I needed 100 widgets to meet 100% of demand. Now, I need only, say, 33 widgets to meet demand, and so I build 33 and call it a day.
But what if I discovered a technology that made my existing 100 widgets 3x more effective? And moreover, every widget in every industry became 3x more effective at the same time?
To put that in perspective — roughly 60% of the US population is employed at any given time. If that same 60% were suddenly to go from producing 100% of our needs to producing 300% of our needs… what does that mean for society?
The first instinct is to think that means we’re in for massive job loss. And in some ways, that may be true — but not in the way that we may think. Remember this is not job reduction due to lack of demand or lack of a viable business model, but because reduction in labor does not reduce the availability of product. Put another way, it would theoretically be possible for 2/3rds of the work force to quit their job and continue to earn a livable wage.
Before you say this sounds like crazy talk, remember that this is the goal of returning to the era of single income households. Instead of two or three people having to work in a household, one person working can produce enough value to support the entire family. Making everything 3x more efficient is not subject to the same inflationary forces as saying you’re going to start paying workers 3x more.
Instead, it’s like saying that you’re going to eliminate 2 of 3 job positions, and at the same time make everything 3x less expensive for everyone.
Is that job loss in the traditional sense? For families, it could mean the same quality of life with 1 full-time position, instead of three. For an individual, it could mean the same quality of life on 1/3rd the hours. While I think it is unlikely to play out that gracefully in the short term, it is possible to have a fundamental shift in how we think about productivity without catastrophic consequences.
But here’s the additional wrench in that thinking: I don’t believe that we’ll see a 2x or 3x improvement in efficiency and productivity through AI. If you look at the speed at which AI is currently increasing the rate of production, we’re seeing multiples in the thousands, not one or two. A piece of current generative art created in 30 seconds could easily take a human artist 50 hours to create — a multiplier of 6,000 times. And that leaves out the things that simply are not possible with today’s level of communication — the exponential connections that would never have happened at all. It may be reasonable to imagine that two people living in the same city in the 1500s could meet each other through chance, but it would be unreasonable to assume that they could have met each other if living on opposite sides of the world, without either of them ever leaving their home country.
Today we accept as common what was once impossible, not just 2x or 3x improved. A similar level of advancement in every aspect and industry in our lives at the same time will make the miraculous commonplace.
What I think we’re not discussing is how much that level of optimization will impact life below our conscious level, through personalized AI as individual actors talking among each other for a positive mutual outcome. Is it 2x or 3x as we discussed above? Is it 10x? 100x? 1,000x?
At 10x, 600 people could support a town of 10,000. At 100x, only 6 people would be needed to support a city of 10,000 at today’s standard of living.
It’s either stupid magical thinking, or entirely predictable.
The Future of Emerging AI
The rate of AI development is so rapid at this point that it’s really difficult to conceptualize in advance what it will impact next. To reference the harnessing of electricity again — once it was discovered, the world went through a period of doing everything we were already doing before, but upgraded with electricity. Pick any aspect of life and ask how it could be improved with electricity. Brushing your teeth? Taking a shower? Getting to work? Doing laundry? Exercising? Cooking? Socializing? Making love? Pick an aspect, add a bit of electric shock, and you have a product.
I think we’ll see the same with AI. We’re some distance from a true general AI that can do what we’re imagining, but as we get closer, we’re seeing thousands of applications around v.1 of AI come out the same way that electricity birthed gadgets for every aspect of our lives.
But if you’re looking for the more subtle places that others have not touched on as much, I think I’d look for every interaction where probability plays a substantial role. Find fields with what was once “acceptable tolerance” for randomness, an amount that we’ve overlooked because it’s too small to worry about. Reimagine those situations with what I call “biased probability”. In those places, I believe you will find the magic of magical AI.
A Final Thought on Perfect Knowledge
I think everyone has imagined at some point what superpower they’d want if superpowers were real. Turning invisible? Super strength? The ability to fly?
At some point, I decided that I wanted the “Power of Next Immediate Action.” I wouldn’t want to know the future — because that wouldn’t be any fun — but I’d like the ability to know, magically, exactly what action I should take next to accomplish my goal. Why build a missile to destroy Superman if all you had to do was row out into the ocean and drop a pebble of Kryptonite at just the right spot, and a year later he chokes while eating sea bass?
Putting aside that I apparently envisioned myself as the super villain, for some reason, I think it’s an apt analogy for what is coming. We keep talking about AI as if it’s super strength, laser beams from the eyes, and super speed. And it is those things. But the far more powerful and more subtle effect on everyday life will start simple and small, and most people are not yet thinking this way.
It will be the age of the lucky. Lucky day after lucky day. You’ll have umbrellas handed to you moments before the rain starts, and you’ll be asked to hand your umbrella to someone leaving the building as you arrive. Food will arrive right when you start to get hungry, at just the right portions. Your life will become easier.
And by then, it’ll be just a minute to midnight.
