You Probably Don't Value Your Time, Money, or Energy
What would the stoics say about how people use AI? People are suboptimal time allocators.
I’ve been noticing a big problem lately amongst people and how they relate to money, technology, and time.
We are essentially in an exponential technological takeoff right now, but people kind-of-sort-of don’t want to admit it.
And I don’t think many people are willing to evaluate what it means for their day-to-day lives, or how to use these new tools.
But beyond that, I don’t believe most people have taken a serious inventory of how they spend their time, or how much their time is worth.
I say this because I frequently try to convince friends who work in technology to try and use GPT-4. The main problem for them is that it costs money: A whopping $20 a month.
$20-fucking-dollars a month.
“Great, another subscription that I’m never going to use and forget to cancel” they think to themselves, after this sticker shock.
It really doesn’t matter to them how many benefits I list off, or incredible examples of how I use it every day and how they could too.
At the end of the day, it’s another $20 for what’s probably just another technology cycle hype-fad.
What’s funny to me about this is that people generally have absolutely no issue with wasting $20 on something relatively contrived or useless. They will spend that much on two beers at a bar; They will spend that much on Netflix.
But as soon as it comes to paying $20 for a literal tireless intelligence that can save you hours of work and energy each day, it’s time to batten down the hatches.
What this tells me, however, is that people really don’t value their time, and I think it’s because they don’t consciously realize what their time is worth, and how they spend it.
Conservation of Time
Let’s look at time spent on a typical coding task before GPT-4:
A developer receives a task to implement a function. Let’s say the task requires using a new library the developer has never used before. Typically, the developer would google something like:
”How to use [library X] to implement [requirement Y]”
The developer lands on the library’s documentation page. Unfortunately, the documentation lacks the specific example they’re looking for.
The developer then searches StackOverflow (the premier Q&A website for coders). StackOverflow has an answer, but it’s outdated and uses an older version of the library, so the answer doesn’t work.
Sometimes someone posts the updated answer far down in the comments. Otherwise, the developer often has to dig into the source code of the library, find the method, and figure out how to use the method from scratch.
After this, they must then synthesize how to use the library and methods for their exact use case.
That’s a lot, right? This whole process can take half an hour, or longer.
And then, here’s the workflow with GPT-4:
Developer types “Write a method that implements [requirement Y} using [library X].”
And then GPT generates the source code. Probably close to 8 times out of 10, the code can be pasted into the code editor and used immediately. And about half the time for the few times it doesn’t work, the error can be pasted back into GPT and a working version of the code is returned.
This process might take 3-5 minutes.
Now let’s say the average US developer makes roughly $52 an hour.
And let’s say this massive time-saving workflow that I just described only happens once per month saving the developer 25 whole minutes.
At their hourly rate, the developer would have saved $21 worth of their time!
So even if this flow only happens once per month, the developer is essentially paid $1 to save a significant amount of their time.
However. this flow happens constantly. I’m assuming this because since ChatGPT’s release, StackOverflow usage has precipitously dropped 35%, sinking from 16 million daily page views to only 9.6 million.
However, I’ll say personally that I use it at least a dozen, if not up to 50 times a day, and I know that it’s saving me a ton of time and mental energy.
Beyond that, people don’t even consider the hundreds of other ways it can save time. For instance, one can describe a workflow to GPT, and it can use plugins to create visual diagrams of the workflow that can be shown to others within minutes, when it can take much longer to build the diagram by hand visually.
In fact, here’s a diagram of the developer workflow I described above. I just pasted it into ChatGPT, tagged the separate processes, and got this in 30 seconds:
Resisting Change
Here’s the thing: You can almost never go wrong by over-indexing your attention on new technological advancements.
The people who were super jazzed about books after the Gutenberg Press came about eventually became highly educated aristocrats and their protégés helped bring about the enlightenment.
Those who became familiar with factories and tooling figured out how to improve efficiency of assembly lines and sell tools that hastened assembly work, while factories and workers who refused to use the new tools lagged behind.
And those who could figure out how to program, use computers, and leverage new software could perform leaps and bounds more quickly than people who continued to perform calculations and crunch numbers by hand.
In essence, learning and using new tools that have a low barrier to entry is often very much worth it for the tool user! However, people quickly adapt to their current circumstance, forget the big picture, and give up trying to use new tools as soon as they become comfortable with their current process.
I want to stress that it is not just tech workers that suffer from this problem, though they are the most egregious offenders because they should know better.
99 times out 100 when I talk to people who have used ChatGPT, they have not used the paid version.
Most college kids and school students have used the free GPT to generate essays, but that’s the total extent of their usage. Despite it being a tireless tutor in almost any subject, the paid version is just not viewed as “worth getting”.
Conservation of Energy
Why do executives hire assistants? Usually, it’s to deal with the minutia of day-to-day operations, and reduce the number of decisions they need to make so they can focus on the biggest decisions of the day.
This is one of the bigger benefits AI also gives to us. It makes smaller decisions for us, so we can remain focused on the bigger picture, and act as an orchestrator of many moving parts.
It’s a similar principle to how it’s easier to edit an essay than it is to start writing one from scratch. If GPT generates a code function, or a diagram, or a PowerPoint presentation, it’s usually mentally easier to start editing that, than it is to try to create it from scratch.
And that conservation of mental energy is worth something significant as well – it enables you to make better decisions when it counts, rather than face overwhelm from every small decision you had to make in the day.
Conservation of Money
Another funny aspect about this is I don’t know a single person who, once they started paying for GPT-4, ever went back to the free version. It instantly changed their workflows and made their lives easier. Many people don’t know how they lived without it, and wished they would have had it 5 years ago.
I’ve even seen people mention that they would willingly pay $500 a month for what GPT-4.
And that’s kind of the crux of it. If you have a tireless, instantaneous intelligence available for you at your beck and call, you probably could figure out a way for it to make you some money, right? Especially so if it enables you to finish work 2x-5x as quickly as others.
And, even if it only saved you time, rather than made you money, I think it’s still worth it. I think people should price their time higher. I believe that everyone should price their personal time at $5,000 an hour (on an emotional level). Anything that can give you back an hour every day is worth it.
Every hour saved on menial tasks is another hour that can be spent on further learning, or exercise, or spiritual practice, or time spent with family. Time is worth a lot – certainly worth a lot more than money.
But it’s also worth noting we also have not yet reached an upper-bound on the return-on-investment from using AI to do our work. As these models become smarter, we’ll likely see even more resources being poured into them to get time, money, and energy given straight back.
On a related note, I think it’s important for people to stop sneering at the sticker price when paying for additional AI tools built on top of GPT.
For instance, I recently mentioned an AI code-editor plugin that could generate unit tests, catch bugs in pull requests, and automatically recommend and improve the entire codebase to some developers. They felt very leery about the notion that the plugin cost $20 a month.
However, as I illustrated before, if a $20 tool saves a developer 25 minutes per month, it will have paid itself off.
And when you consider that shipping a single bug to production often takes hours to fix, and costs hundreds, or even thousands of dollars in customer support, development labor, and reputation damage — a tool that can automatically minimize bugs suddenly seems like a bargain.
Furthermore, anyone who works on independent projects, or freelances, is able to take on and earn a lot more with the capabilities this time enables.
The bottom line is that people are far too cheap with their very expensive time.
"People are frugal in guarding their personal property; but as soon as it comes to squandering time, they are most wasteful of the one thing in which it is right to be stingy." - Seneca