Algorithms for Life: Letting Go

Welcome to Udemy Research from our Algorithms for Life series by Valor Ingles. Okay, so imagine the scene. It's April 13th, 1970. You are 200,000 miles from Earth. And suddenly, there's a bang. An oxygen tank has exploded. The Odyssey spacecraft is crippled, power is dying, the carbon dioxide levels are rising fast, and you and your two crewmates are going to suffocate. Why? Because the square filters from the command module won't fit into the round openings in the lunar module where you're, you know, hiding out. It's the literal definition of a square peg in a round hole problem. Exactly, but the stakes are life or death. Right. And back in Houston, flight director Jean Kranz is staring at his team of, uh, terrified engineers. And he issues an order. And I think it's one of the most important sentences ever spoken in the history of problem solving. He says, I don't care what anything was designed to do. I care about what it can do. And that just shifted the entire paradigm. The engineers stop looking at hoses and plastic bags and duct tape as hoses and bags and just saw raw material. And they built the famous mailbox adapter. The CO2 levels dropped and Apollo 13 came home. Yeah. Now we usually tell that story as a triumph of human grit, right? The ultimate McGiver moment. But what if I told you that Kranz wasn't just improvising? He was, uh, actually performing a rigorous computer science operation called constraint relaxation. That is exactly right. This isn't just about duct tape. This is about a fundamental algorithmic strategy in computer science. When you face a problem that is for all intents and purposes impossible to solve the time you have, the only way forward is to temporarily break the rules. And that is what we are unpacking in this deep dive. We're looking at two counterintuitive strategies from computer science relaxation, which is letting go of constraints and randomness, which is letting go of control. And the evidence suggests that for most of us, our biggest failure mode is that we are just too rigid. We follow rules that don't exist and we avoid changes that would probably make us happier. But, and this is a huge butt that we're going to get to later. There is a trap here because the same principle that saved Apollo 13 is what caused the Boeing 737 MX disasters. So the mission today is to figure out when to let go and when holding on will save your life. Let's get into the foundation of this. You mentioned impossible problems. I mean, I have a hard time deciding what to watch on Netflix. But what is a computer scientist consider an impossible problem? So in CS, we call these NP hard problems. Let's take a classic example. Placing security cameras. Imagine you have a building with 10 hallway intersections. You want to place the minimum number of cameras to cover every single hallway with 10 intersections. A computer can check every combination instantly. It's easy. Okay, 10 is easy. What about 100? With just 100 intersections, the number of possible combinations exceeds the number of atoms in the observable universe. Wait, really? Just 100 intersections gets you to more than the atoms in the universe. Exponential growth is terrifying. So you cannot just check every option. It would take longer than the age of the universe. You cannot optimize this perfectly. It is effectively impossible. So what do you do? Do you just guess? You cheat. Well, mathematically speaking, you use a technique called linear programming relaxation. See, in the real world, a camera has to be either there or not there. It's binary. One or zero. That's a hard constraint. But in relaxation, you tell the computer, okay, you can put half a camera here. You can put 0.3 cameras there. Which is useless in reality. I can't go to Best Buy and Buy 0.3 of a Nikon. Precisely. It sounds absurd. But by relaxing that integer constraint, the math becomes incredibly easy. The computer solves the relaxed version in millisecond. Okay, but hold on. If the computer comes back and says, put half a camera in intersection A, and then I have to round that up to a whole camera to make it exist in reality. Haven't I just messed up the optimization I'm adding costs back in? It feels like I'm back to guessing. That's the intuition. But the math says otherwise. This is the magic of the bound. Because you calculated the perfect fractional solution first, you know the absolute mathematical floor of the cost. When you round up, the math proves that your solution is guaranteed to be within a specific margin. Say, it's never going to be more than two times the perfect score. You aren't guessing blindly anymore. You are tethered to the perfect answer. You accept a slightly imperfect real world solution because it's the only way to solve the problem before the sun burns out. Exactly. I see. So you break reality to fix reality. And once you start looking for this constraint relaxation pattern, you see it everywhere in history, not just in code. Absolutely. Look at the Voyager 2 mission in 1965. Gary Flandro, a JPL had an impossible constraint. Getting to Neptune would take 30 years. The funding and the technology, they just weren't there for a 30-year mission. So the hard constraint was spacecraft have to fly in a direct path. Right. Flandro relaxed that. He realized if he used a rare alignment of the planets, something that only happens once every 175 years, he could use gravity assists. He broke the direct path rule. And the result. They got to Neptune in 12 years, not 30. And then there's the James Webb Space Telescope. I was reading about this. The mirror needed to be six and a half meters wide to see back to the big bang. But the biggest rocket fairing was only four and a half meters. A physical impossibility, a hard constraint. Until they relaxed the constraint of a solid mirror, they made it out of 18 hexagonal segments they could fold up like origami. The essentially said, the mirror doesn't have to be one piece. And this happens in business too. The most successful startups are almost always a result of relaxing a constraint about identity. Oh, like Slack. Stewart Butterfield spent three and a half years building a video game called Glitch. And it was a disaster. The game was a total bus. Right. But they had built this cool little internal chat tool to talk to each other while coding. Exactly. The constraint was we are a gaming company. When they relaxed that and asked, what if we're a communication company? They sold that internal tool to Salesforce for $27.7 billion. Same with YouTube, right? It started as a video dating site. They even offered women 20 bucks to upload videos introducing themselves. And nobody did it. Total failure. But when they relaxed the dating constraint and just said, upload whatever you want. They got the me at the zoo video. And 18 months later, Google bought them for 1.65 billion. Yeah. Okay. So we've established that algorithms, astronauts and billionaires succeed by letting go of the rules. But let's bring this down to earth. Most of our listeners are building space telescopes. They're trying to decide if they should quit their job or move to a new city. Does this math apply to us? It does. But it highlights a flaw in our human operating systems. In algorithms, we use a technique called simulated annealing. It mimics how you cool molten metal. To get the atoms to line up in a perfect crystal, you have to keep the moving. If you cool it too fast, they get stuck in a weird brittle structure. A local optimum. You're stuck on a little hill. But you can't see the massive mountain right next to you because you'd have to go down into the valley first. Exactly. To escape a local optimum, the algorithm sometimes accepts a worse solution. It deliberately goes downhill to find a higher peak later. But humans, we hate going downhill. We are terrified of it. We cling to the status quo. We do. And there is a fascinating study from 2020 by Stephen Lebit, the Frekenomics economist that proves just how stuck we get. He found 22,500 people who are genuinely agonizing over a major life decision. Should I divorce? Should I quit my job? They had been debating it for months. So these are people who are truly on the fence, 50, 50. Right. So Lebit had them flip a virtual coin. Heads, you make the change, tails, you stick with the status quo. Okay. I have to stop you there. This is a marriage we're talking about, not a dinner order. Flipping a coin sounds less like science and more like a nervous breakdown. You're telling me Lebit actually advised people to end a marriage based on a nickel. I know it does sound reckless, but you have to look at who was flipping the coin. These weren't people who were happy. These are people who had been agonizing for months, maybe years. They were already stuck in a misery loop. So the coin didn't make the decision. The coin just pushed them off the ledge they were already standing on. Exactly. And look at the data. The people who got heads and made the change were on average 2.2 points happier on a 10 point scale six months later. Two points is huge. That's like going from a six to an eight. It is massive. And the lesson is that we suffer from severe status quo bias. If you are truly 50, 50 on a decision, you are not actually 50, 50. You are likely staring at a better option, but your fear of the dip that downhill moment is holding you back. That dip is real though. We see it in the j curve data for careers and even divorce. Things get messy for a bit before they get better. They do. But the recovery is faster than we think. And this applies to entrepreneurship too. There is this myth that you have to be 22 to start a company, but the NBER analyzed 2.7 million founders. The average age of a high growth founder is 45. A 50 year old is 1.8 times more likely to succeed than a 30 year old. That is comforting for those of us who aren't wearing hoodies anymore. But I want to push back on this letting go idea for a second. We're talking about satisfying, finding a good enough solution rather than the perfect one. But there's research showing that maximizers, people who obsess over the perfect choice actually get better outcomes. Specifically, they get 20% higher starting salaries. That is true. Well, isn't good enough just an excuse for mediocrity then? If I can get 20% more money by being obsessive, why shouldn't I? I can buy a lot of happiness with a 20% raise. I can buy a jet ski. Have you ever seen a sad person on a jet ski? I have not. That is a fair point. But I've seen plenty of miserable rich people. And that's what the data shows. This comes from Barry Schwartz's research. Yes, the maximizers get the money. But the correlation between maximizing and regret is huge over 0.5. Meaning they get the job. But they lie awake at night wondering if they could have done better. Exactly. They are tortured by the counterfactuals. So yes, you get the objective win the salary, but you get a subjective loss. You enjoy the money less because you're too busy optimizing. Optimization is great for building bridges. It is toxic for happiness. Okay, I'll grate you that. But what about the jam study? You know, the famous one where they put out 24 jams and nobody bought anything because they're paralyzed by choice. If we just relax all constraints and have infinite options, don't we just freeze up? I'm glad you brought that up because we need to correct the record on the jam study. Everyone cites it to say choice is bad. But a meta analysis in 2010 by Shai Bihin looked at 50 different experiments and found the a focus size was virtually zero zero. So choice overload isn't real. It's not a universal law. It only happens under specific conditions. When you have time pressure, the options are complex and you don't know what you want. If you know you like strawberry jam, 50 options doesn't paralyze you. You just grab the strawberry. Okay, so we don't need to artificially restrict our choices just to function. But we do get tired of making decisions, right? Decision fatigue. I've always heard that's because your blood sugar drops. You run out of glucose. I usually eat a snickers before a big meeting. That is another myth we need to bust. The glucose depletion model has been largely refuted by recent replication efforts, specifically the Hager 2016 report. It's not that your fuel tank is empty. It's a motivation shift. Your brain switches from half to mode to want to mode. So eating a snickers is just a placebo. Worse, it's a distraction. If you want to fix decision fatigue, don't feed the brain, change the context. The constraint isn't biological energy. It's attention. Go to a different room, switch tasks. This actually changes everything about how I planned my day. I always thought I was running out of fuel. You're saying I'm just bored. Effectively yes. Brutal. Okay, so we've covered relaxing constraints in the science of decision making. But there's one more piece of the letting go puzzle. Randomness. Serendipity. Yes. We try to plan our careers perfectly, but the data shows that engineered serendipity is a real strategy. This brings us to the weak ties theory, right? Correct. A massive study on LinkedIn in 2022, 20 million users confirm that your close friends are not the best source for new jobs. They know the same people you know. It's the weak ties acquaintances you haven't spoken to in six months who bridge you into new networks. And proximity matters too. MIT found that just being in the same building increases collaboration by 33%. Being on the same floor increases it by 57%. So bumping into people at the coffee machine isn't wasting time. It's an algorithm for innovation. Exactly. If you've spent the last 10 minutes bashing rules and constraints, we've praised NASA for using duct tape and startups for pivoting. But I want to pause here because if I'm getting on a plane tomorrow and I hear the pilot is relaxing constraints and experimenting with randomness, I am getting off that plane. And you should because this is where the Silicon Valley mindset becomes lethal. We're talking about the Boeing 737MAX. We are. And to understand that tragedy, we have to look at it through this exact algorithmic lens. Boeing applied a startup strategy to a safety critical hardware problem. They relaxed the wrong constraints. They wanted to compete with Airbus without the cost of designing a new plane. Right. So they relaxed the engineering safety constraints. They used software MCS to patch over aerodynamic instability. And crucially, they relaxed the constraint of transparency. They hid the system from the pilots. They're prioritized cost and speed over safety. And 346 people died. So how do we reconcile this? We have lev it saying take the leap and Boeing showing us that leaping can be fatal. It feels like a massive contradiction. It's not a contradiction. It's an asymmetry. I call it the asymmetry framework. You have to look at who is making the decision and what the costs are. Individuals you and me are systematically too cautious. We suffer from loss of version. We need to relax constraints to grow. But institutions. Institutions are systematically too reckless. They suffer from moral hazard. The decision makers don't bear the physical risk. So individuals need to be pushed to let go. Institutions need to be forced to hold on. Precisely. If you are in a complex domain like your career or dating, you need to experiment. If you are in a safety critical domain like aviation or medicine, you stick to the rules. Okay. That is a crucial distinction. Let's land this plane safely with some practical protocols. If a listener is thinking, okay, I'm stuck. I need to apply this. What do they do? I have five protocols for you. Let's hear them. Number one. Number one is the calibrated coin flip. This is for when you are genuinely stuck between two options and have been agonizing for weeks. If the coin says change, commit to it for a set period, say six months. And again, the magic isn't the coin. It's that the coin forces you to confront your true preference. If it lands on change and you feel relieved, you know the answer. Exactly. Number two is the constraint inventory. This comes from the theory of constraints. List all the reasons you can't do something, then label them physics, legal or self-imposed. I can't fly is physics. I can't rob a bank is legal. I can't change careers because I'm 40 is self-imposed. And most of them are self-imposed. Test one, just one. Number three. Strategic, satisfying. Before you start looking for an apartment or a job right down your criteria must be under $2,000, must have a balcony. The moment you find the first one that meets those criteria, take it. Stop looking. Believe the app. Don't look back. I like it. Number four. Engineer serendipity. Use the weak tie rule. Once a week, reach out to one person you haven't spoken to in six months. Just a hello. You are widening your surface area for luck. And number five. Environment design. Since we know decision fatigue is about motivation, not glucose, don't try to power through the candy bar. Change the environment. Use defaults. Auto and roll and savings. Make the right choice. The lazy choice. I love that. Make the right choice. The lazy choice. It's the only way I get anything done. So if we pull this all together, the lesson isn't always break the rules or always follow them. No, it's about knowing which constraints are load bearing. Gene Crems didn't say I don't care about physics. He didn't say I don't care about oxygen levels. He respected the constraints that mattered survival. He relaxed the constraint that didn't purpose. He didn't care what the hose was designed to do. He cared what it could do. And for our listeners, that is the question. Which of the rules running your life are laws of physics and which ones are just expectations? Something to think about before you flip that coin. You can find the full stack of research, including the Levit study and the analysis of the 737 MX at research.u to dot me. That's yud a dot me. And remember, sometimes the scariest move is the safest one. I'm Valorangles. See you in the next deep dive.