It’s a basic problem in technology, that if something is totally unlike anything which came before it, then it will be difficult to accurately describe it. People often opt for analogies, and model it after what they know. Unfortunately, these “metaphors” for the project can often take on a certain mythology which obfuscates the actual value of the product.
Chess
Chess is a cool battle simulator game. You have the king and queen, then the pawn is a foot soldier, the knight is a cavalry unit, the bishop was originally an elephant, and the rook was originally a chariot. Maybe this explains the gender imbalance of the game: boys are more into war stuff?
There’s only one problem: no one would ever describe Chess as a battle simulator.
Judged as a battle simulator, chess is terrible for many reasons. In actual battles, all of the units move together on the same timetable, whereas in chess, you only select one piece per turn, and all all the other pieces stay still. Furthermore, the speed dynamics don’t make any sense in Chess. A bishop can move 7 squares in the same number of turns as it can move 1 square? Units aren’t supposed to be infinitely fast! And what’s up with the movement of pieces? In real life, different units may move slightly differently, but the real difference comes from their attacks. In Chess, all the pieces attack the same way (except for pawns, don’t even get me started), and the attacking mechanics are unrealistic to boot!
It’s easy to look at chess, scoff, and say “what a stupid game. If I wanted a battle simulator, I would go to the Total War series, or one of those WWII simulators, or one of any number of games that let me manage an army.”
But you would be missing out. By no means is Chess “realistic”, but who said it was supposed to be? It can be enjoyed if you take it for what it is: a strategic turn-based game with complete board information, no RNG, and a near-infinite amount of tactical depth.
Social Media Companies
Many successful software products (most?) were conceived for a different purpose than they ended up being used for. This can be thought of as a form of “pivoting” – you design the product to do X, but you quickly realize that it’s much more useful for Y, so you pivot to Y.
Apparently, YouTube was originally conceived as an online dating site, before moving to online video. But example doesn’t excite me too much, because the founders moved away from that idea very early.
I’m more interested in the example of Twitter, which was conceived as a “micro-blogging” platform. The idea was for people to post short updates of what they were doing throughout the day- stuff like “I’m having toast for breakfast”, or “I’m at the park”. Or maybe that’s the strawman version of what people thought it was for? Either way, the product ended up being… not that.
These days, all social media websites end up looking the same: an infinite-scroll “feed” of either short-form-video or multimedia content. It’s the homogenization of media, with the mythology of “140 characters” or “making friends” or “TikTok dance trends” on there as historical artifacts which inform the culture of the sites.
Bitcoin
When the Bitcoin source code was first posted to Hacker news, on May 8, 2009, this was the comment that was posted:
Well this is an exceptionally cute idea, but there is absolutely no way that anyone is going to have any faith in this currency.
On May 2009, the price of Bitcoin was approximately zero. As of my writing this post, the price of Bitcoin is approximately 63 thousand USD.
Suppose you were browsing Hacker New circa 2009. You see a post with 5 upvotes touting a potential digital currency. You think for a second, maybe you even look at the source code. You decide that it will not be viable as a currency.
You would be correct. In the United States, Bitcoin is not a widely used medium of exchange. The commenter wasn’t wrong.
Cue the Bitcoin enthusiasts fighting me tooth and nail on this point. Look, I know there are certain vendors that accept Bitcoin. Maybe you even purchased something with Bitcoin yourself. But let’s be honest, when was the last time you saw the person in front of you in line at the grocery store pay for their food with Bitcoin? Bitcoin purchases are a novelty. There are other cryptocurrencies which get some traffic as mediums of exchange, but only for niche use cases, where there’s some exceptional reason to explicitly circumvent the legal tender.
Arguably, the problems with Bitcoin as a medium of exchange are baked into its design. People expect their money to experience a steady ~0-3% inflation YoY, not massive swings of appreciation and depreciation. Bitcoin enthusiasts celebrate when massive appreciation, they are unknowingly discrediting it as a currency, because it’s bad when a currency appreciates in value – that’s a phenomenon called deflation.
So Bitcoin failed as money? The only problem for you as a 2009 Hacker News reader is, by not investing, you missed the best opportunity of your life. Imagine that: you were exactly in the right place at the right time, one of ~5 people who happened to see that post, on the one website likely to post about Bitcoin, and you missed it.
This point resonates with me, because I browse Hacker News, partly in the mindset of an investor, looking for interesting projects to dig my teeth into. I just need to remember that “interesting project” does not equate to “I agree with the description of the creator.”
Whereas Bitcoin failed to usurp the US dollar (the propaganda answer to “what is bitcoin?”), it succeeded at an entirely different task: becoming a competitor to gold. Both Bitcoin and gold are investment assets with financial value rather than intrinsic value (the practical uses of gold notwithstanding).
Neural Networks
Sometimes, someone will claim that neural networks work so well because they operator according to the same principles at play in the brain. That’s why they’re called “neural networks” – when you graph it out, it resembles neurons and synapses.
Then someone else will jump in, and say that actually, it’s not clear that digital neural networks work at all similarly to biological neural networks. Minds are really much more complex, we aren’t close to understanding the inner depth of the human brain, and also, aren’t you a hubristic tech bro for even suggesting the similarity?
Then the first person will go back and amend their claim. Biological and digital neural networks are not the same, but at very least, the former inspired the ladder. Maybe you think this is a case of biomimicry, or maybe the “inspiration” only extends to the choice of name for the technology, “neural network”. Both possibilities are encompassed by the term “inspired”.
And then the debate launches into the the connection between brains and computers, whether that is even an appropriate analogy, whether brains have magical properties which set them apart, whatever.
Obviously, I come out on a particular side of this debate. That the analogy is fine, and that the neuroscience people are being unnecessarily thick by rejecting it. On the other side of that, I think that computer scientists should put more work than they do on experimenting with architectures which are more similar to brains then our existing models. But that’s neither here nor there.
Ultimately, though, this debate has no bearing on short-term discussions of capabilities. Neural networks based on back propagation have proven themselves to be extraordinarily powerful. If you reject digital neural networks for the superficial reason that they differ from those in brains, then you’re turning a blind eye to their ability. ML models can undeniably do things which, up until now, only brains could do. Getting bogged down in whether or not we can claim the underlying algorithms are similar doesn’t change that fact.
So What?
Maybe it’s clear to you what all of these stories have in common, or maybe you’re aren’t sure. What I’m trying to do here is refine one of my mental models. Sometimes, it is useful for me to adopt the mental model of an investor. “Would I invest in this project? What is the risk, what is the upside, what is it’s scalability?”
This mental model has a superstitious failure state. I like to form opinions about whether people are right or wrong. So when I come to the conclusion that the founder of a project is wrong about something, then I discredit the project. When that happens, I need to catch myself. It’s useful to know whether the founder is correct, but if I assign that full importance, I’m playing the wrong game.
If I don’t invest in chess, because it’s not a good battle simulator, I’m missing out on fun. If I don’t invest social media companies, because they’re based on dumb concepts, I’m missing out on profit. If I don’t invest in bitcoin, because I’m never going to use it as money, then I lose that opportunity. And if I dismiss neural networks, because they’re not like brains enough to satisfy me, then I’m not focusing on the most promising technological development of our lifetime.
Epiphany a Week 