Every day we see our leaders and experts making mistakes. If only they put me in charge, we say, I could be so much better. This kind of hubris is known as the Dunning-Kruger Effect. The less we know about something the more confident we are in our solution.
But there is also a countervailing force: sometimes the leaders and experts ARE wrong. So the question is, when are you more likely to be right than the experts?
Is it…
Stock picking?
Finding alternative health solutions?
Starting a business all the experts say will never work?
In this essay I’ll show why beating the experts on stock picking is nearly impossible, finding alternative health solutions is tricky but doable and startups that prove the experts wrong are the norm.
For more detailed analysis, read Inadequate Equilibria by Eliezer Yudkowsky, where many of these examples are taken from.
An analogy
Imagine a world where people are very sensitive to the sun—the higher elevation they are the worse off they are. So they try to live at as low elevation as possible. And these are people with no access to the outside world or GPS satellites or anything like that.
So they try to go to some place with lower elevation but every direction they go has higher elevation. They conclude this is the lowest place in the land.
But one day a brave soul, Alex, says “I’m going to find a lower place for my people”
“No!” the people cry. “We have already looked and this is the lowest place. You are going to burn up!”
Despite this, our hero (or village idiot, as Alex is now popularly known) starts climbing the hill. Alex finds that eventually the land starts going down and finds a new place to live that is further from the sun. Wow, Alex thinks, won’t my people be so happy that a found a new place to live that is better?
Alex braves the mountain peak to go back to bring the good news. But gets an unexpected response. “You expect us to believe that there is a lower, better place than our dear homeland?! Most certainly not!” Rather than admit they were wrong and update their opinion they’d rather blame someone else for their ignorance.
This example shows not only how hard it is for the “experts” to update their point of view but also the kinds of situations where one single hardworking non-expert is likely to find a better solution.
Real-world examples:
Internal combustion engine (ICE) cars are pretty great, right? And they have vastly improved since they were invented in ~1900.
But there was always a feeling that cars could be better. For example, ICE cars are pretty inefficient. About 17% efficient. Meanwhile, an electric motor would be about 70% efficient. Given that electric power plants are around 50% efficient electric cars are still ~35% efficient, twice as efficient as ICEs. Electric cars would have other benefits like low smog and low noise.
But there was no clear way to get there. ICE cars are cheap and the barriers seemed insurmountable. All attempts at electric cars either failed (the GM EV1) or were not true electric cars (Prius Hybrid).
As Yudkowsky says in Inadequate Equilibria, getting over one of these hurdles, is so expensive that it requires someone with a lot of money (billionaire level), strong conviction and little concern for losing their reputation (or their money) as they are likely to be proved wrong and ridiculed. Such a trifecta is rare in an individual.
Lucky for us mere mortals such a person did exist (otherwise it might have taken much longer to get electric cars).
Sidenote: This is why taxing billionaires has downsides for society. Better to encourage them to invest in such world-changing technologies like electric cars or better healthcare. We kinda do this by making non-profit donations tax-exempt but non-profits rarely create such world-altering technology as Tesla or SpaceX, which is providing uncensorable internet to the most remote villages in the world through their internet service, StarLink. Thus, efforts to Tax the Rich has the unintended side effect of depriving the world of world changing technologies. Instead of Tax the Rich we should Incentivize the Rich to invest their capital in potentially world changing ideas.
The reality is that all new businesses have some kind of activation energy that have to be overcome. Some activation energies are higher than others. For example, though creating Facebook was no trivial matter, it wasn’t as hard as starting an electric car company— it could be created by college students, evidently.
Meanwhile, other areas are much harder. Let’s take healthcare. Somehow the US has the world’s most expensive healthcare per capita and without the best results. And anyone who has been to a hospital has the distinct impression that there is room for improvement. You don’t need to be an expert to have the gut feeling that something is wrong here.
But no one seems to be able to fix it. Sure, there are lots of theories but for whatever reason no billionaire has fixed it. Even the effort by Amazon, Berkshire Hathaway and Chase Bank to make something better with a world-famous surgeon at the helm failed.
So healthcare looks something like this:
Some general principles:
Any time the world has reached an equilibrium that is not optimal we can say it is inefficient or inadequate.
There can be many reasons for inefficient systems. In the 2000s, Japan experienced declining Nominal GDP. A primary reason, speculated by some economists was *low* inflation rate. If only the money supply was increased, there would be more investment, people would spend more, there would be more jobs and eventually more exports too. So why didn’t the central bank of Japan change things?
The central bankers were disincentivized from solving the problem. Here’s why: central bankers want to keep inflation low. No one wants to be the next Weimar Republic (Germany pre-WWII) or Zimbabwe or Venezuela, right? So central bankers are left with a situation where if the economy improves they are heroes, regardless of inflation. If the economy stays bad and inflation is low, no one can fault them. But if the economy stays bad even as inflation rises they will lose their jobs. It’s like this:
President Abe ran on the platform of increasing inflation and when he did increase inflation the economy seemed to do better too (though hard to tell what caused what in such a complex system).
Even though many people thought Japan’s inflation was too low, there was no way to exploit this inefficiency. In other words it was unexploitable.
On the opposite end of the exploitability spectrum is stock trading. Exploitability explains why a layman contrarian shouldn’t expect to beat the stock market. Inefficiencies in stock pricing are easily exploitable by any of the tens of thousands of people staring at computer in financial centers and at home. And they can do it on their own without needing to form a whole team or solve some other complicated problems first—they buy or sell a stock.
Do you know what the most effective treatment for depression and anxiety is? Exercise. Yes, Exercise is more effective and longer lasting than pills for treating depression. And there are many studies on it so it seems repeatable. So why do we hear drug ads on tv for depression and not exercise ads for depression? When I've gone for therapy no psychiatrist has ever suggested exercise. Maybe because it's hard for large corporations to make money prescribing exercise. … gyms obviously make money but they never advertise the anti-depression effects. It's unexploitable.
Collective action problems are different. Many people need to get together to solve these problems. I work with an entrepreneur in Kenya who has developed an emergency hotline that gives world-class service to hundreds of thousands of Kenyans and hopefully scales across the developing world. Let’s give a scale for things. In the US it takes about 10 minutes to get emergency medical help if you call 911. In Kenya, it used to take 164 minutes to get medical attention. With Flare, despite the poor infrastructure and traffic jams, they have reduced the time to get medical attention to 18 minutes on average, just 3 years after launch. How did they do it?
Emergency healthcare is a collective action problem. In the US, local governments host an emergency number with dispatchers who connect with the nearest available ambulance and the ambulance brings the patient to the nearest available hospital.
The problem in Kenya is that although there is a government emergency number, the number is not connected to any ambulances or hospitals. You can call them, they will take your details but then they will tell you to call the ambulance. There are 200 ish ambulance companies so you can’t know which is the closest one. Further, since people don’t have addresses in Kenya, you can’t tell the ambulance where to go. You’re dead by the time they arrive. So people take matters into their own hands, get a taxi (if they can afford it) and get to the hospital (which may or may not have the facilities to treat them). This is why it takes 164 minutes for patients to get medical treatment in Kenya; they only get treatment once they arrive at the hospital.
You might think, “oh, it’s a poor country, they probably can’t afford ambulances– the solution is to get more ambulances” but Kenya actually has more ambulances per capita than the UK. So what the heck is going on?
So why doesn’t the government just create a functioning emergency hotline? Like central bankers in Japan, the government is disincentivized from doing that. If they pledge in the newspaper to make 911 better they will probably look bad. Things will probably get worse before they get better. If you don’t try you can’t fail, right?
What about ambulances or hospitals solving this problem? It’s good for their business, right? An ambulance company would only want to send their own ambulances and hospitals would only want to send patients to their hospitals.
Some third-party needs to step in.
But how do you get really poor people to pay for it? And even if they could afford to pay for it, how would you get people to walk around with $100 in their pocket so they can pay for the ambulance if they ever have an emergency. (Remember you can’t bill people in Kenya because they don’t have addresses.)
These entrepreneurs came up with a clever solution: employers want their employees to have emergency medical insurance so that if their employees are ever hurt on the job the employers are off the hook. Plus, it’s good for employee retention.
Flare created an employer-funded emergency insurance fund and a dispatch system to stitch together all the ambulance companies and hospitals.
This is an example of Collective Action failure that a non-expert, non-billionaire was able to solve. Collective Action problems are a treasure trove of startup ideas–one clever entrepreneur can unlock financial opportunities that “experts” can’t see because they aren’t incentivized to see them.
Yudkowsky shares another anecdote of coordination failure: this time on light therapy. Some systems are inefficient because they are unscalable or niche. Yudkowsky’s wife has Seasonal Affective Disorder (SAD) which means she gets depressed in the winter. The usual treatment is to sit in front of a lamp for 30 minutes a day during the winter. But this did not help her. The next solution was to move to Chile (or somewhere sunny) during the winter, but this would be expensive and disrupt their life.
Yudkowsky thought “If the issue is sunlight, and one lamp for 30 minutes isn’t enough, let’s just add more light. Has anyone tried that?” He thoroughly researched the subject online and bought hardcopy books that might contain info that might not be accessible online but no one ever mentioned it working nor even that they tried it and it didn’t work. About 4% of people have SAD and half of those have SAD that doen’t respond to normal treatment (1 light bulb, 30 minutes a day). What are the chances that a non-expert, non-doctor who has done thorough research can come up with a medical treatment that actually works?
It’s worth thinking about. People come up with “medical solutions” that don’t work every day. My family loves to use homeopathy. But in 168 out of 168 double-blind studies homeopathy has been shown to be ineffective compared to placebo. So what makes Yudkowsky’s solution more likely to work than homeopathy? It’s a solution that should be obvious to everyone: if a little sun is good, more sun might be better and thus quite obvious.
Answer: 1st principles and coordination failure.
1st Principles: Yudkowsky has looked as deeply at the problem as possible. SAD only happens in the winter and we already know some light helps some people.
Secondly, coordination failure. For lots of light to be prescribed, lots of coordination is required from the stakeholders:
Psychiatrists can only prescribe treatments that are tested.
Researchers don’t want to test this because it doesn’t fit people’s conception of what treatment is.
Entrepreneurs don’t want invest in this because it’s not protectable (any other entrepreneur can create their own collection of light bulbs). Moreover, it doesn’t affect that many people so the market size is small and there won’t be many repeat customers: people need to buy more pills every month, but people can buy LED light bulbs once and they are set for 20 years.
Yudkowsky tried lighting up his house like a Christmas tree and it seems to have worked for his wife! She no longer has SAD.
Medical fields are a treasure trove for beating the experts actually. But also a landscape full of traps. My friend had the strange issue that he lost his voice and no doctor could find the cause. He’s the type to research problems and not give up. He tried different diets and found that avoiding a group of vegetables called nightshades (potatoes, eggplant, peppers and tomatoes) seemed to make his voice come back a little, whereas other diets like gluten-free or lactose-free did not. Further research revealed that these plants have a little bit of what makes Deadly Nightshade deadly, namely solanine. And in some studies, it looked like some people had an allergy to it. His voice came back and now he can talk again! This was amazing.
So why was my friend able to beat the experts? Patients have first-hand knowledge of their bodies and the inputs to their bodies. More importantly, they have the alignment ofincentives to figure out what is going on. It’s their bodies, after all.
But this led right into a trap for my friend. If doctors could be wrong about nightshades, what else were they wrong about? He became convinced that minute amounts of lead in fish were causing him depression. It got to the point that was all he ever talked about and it was hard to be around him. Maybe lead is a problem. But maybe it’s not. That his first discovery that debunks the experts is so similar to the second seems unlikely, considering he has no expertise in human health prior to this.
The extreme case of misaligned incentives is conflict of interests. A lawyer advocating more legal work or a surgeon advocating surgery or a business lobbying for more subsidy is something to be suspicious of. They might be right but they are more likely to be biased.
But, as they say, “no conflict; no interest.” That is, only those with something to gain from the situation take an interest in it. Why would some random person off the street care if you get knee surgery or not? Thus, the opinions of most experts are compromised. This further proves that a non-expert should be more right than an expert in some cases.
Relative expected frequency for different kinds of coordination failures helps predict when a thoughtful, motivated non-expert might discover or create something that experts couldn’t or thought was impossible.
Truly New Discovery. The most rare is to discover something new all on your own. The discovery of light bulbs for instance. Or that x medicine cures y disease. Someone could do this 0 to 2 times in their lifetime
Synthesis for personal use. Yudkowsky does this with lightbulbs to cure SAD for his wife. It’s not clear if this works for anyone else. Plus, it’s merely an extension of existing knowledge. A committed non-expert might do this 1x per year.
Pick the correct side of an existing argument between experts. I don’t have any original research. I’m not synthesizing existing research. I’m just reporting what experts say. Then I explain why I side with one instead of the other. We do this when we get a second opinion when we go to the doctor–we are choosing which of two experts to believe.
Drawing these item on a spectrum highlights that intermediate points are also possible.
Omelette Error
This is a common mistake for young rationalists to make (young entrepreneurs, scientists, etc), Yudkowsky says. It goes like this: many people are trying to make good omelets. But all the omelets are terrible. You know you can make better omelets. So you get to work and after quite a bit of financial and time investment you are making great omelets. But no one buys them. Why? It turns out that the market says it wants great omelets but that’s not actually what it pays for. It pays for undercooked egg-slop that looks good for Instagram photos at trendy brunch venues.
The omelet error happens all the time in my world of grant writing for social impact businesses (and Yudkowsky says it happens all the time in science grants, too). Grantors say they want to fund innovative projects that have the most impact per dollar. So lots of young entrepreneurs try to build first-of-a-kind businesses that are far more impactful than what is out there. But getting the funding is hard. Why?
What grantors really want is prestige, good optics and low risk. Prestige is getting their name or their boss’s name somewhere near “Nobel Peace Prize” or “Harvard” or at least as few layers removed as possible. Good optics means something that makes for a good picture with a good story; building the first tomato canning factory in Kenya might be the most impactful, dollar for dollar, but training farmers in a field surrounded by ripe tomatoes makes for a much better photo op, doesn’t it?
versus
And low risk means the project has been running for at least three years and has a stable budget. Sorry, but if a company has been running for at least 3 years with a stable budget is unlikely to be innovative.
In science grants, people say they want to fund innovation, but when decisions are made by committee and when people’s main objective is to not be fired, they fund whatever was funded before, so we get stuck in a local maxima—an inadequate equilibria.
But how would a young rationalist know they are in such a situation?
Here’s some guidance on when you *might* have a chance of beating the experts:
Examples of contrarian “technologies” that are unlikely to work from XKCD comic:
[Not to be confused with “Companies making a killing selling it to OTHER people who think it works.”] Source: https://xkcd.com/808/
What times have you gone against expert consensus?
Very educative