Ask Me Anything

I think you're absolutely right that the upside of Meta is better. It's hard to tell how the team will be, but I think what you're trying to do right now is build toward a brighter future for yourself.

Amazon on your CV is going to be nowhere near as strong as Meta on your CV. Since it's an internship, there's very low downside. Even if you don't like your team or the work, you at least get it on your CV. Honestly, even the most boring jobs can be made quite interesting if you set your mind to learning as much as you can.

The other thing about Amazon is that it is not anywhere near as focused on developer happiness as any of the other major software firms. To Amazon, I think software engineers ultimately still feel like a cost center. Almost any other major software firm views their developers as their key asset, not a cost to be reduced.

Farmers went from 40% of our population to 4% in 100 years. AI's dominance will happen much, much faster than that — but I don't think any job goes to 0%. The question is whether the job becomes like farming, where millions are still employed, or like carpet weaving, where only artisans charging $10,000 per carpet are employed.

Anyone right now who says they know what's going to happen in AI or in job displacement is likely just guessing. I don't see how anybody could possibly know the answer for sure. But I think there are a few broad positions:

1. Nothing stops AI from continuing to get smarter than the average US worker. In this world, moving into other intellectual work doesn't make much sense either.

   1. I'm doubtful, though, when people recommend you go into plumbing. It's not clear to me how much longer robots won't be able to do plumbing, especially if you've kept up at all with the advance of robots over the past few years.

2. AI gets very good, but not good enough to replace senior engineers. This is a world where AI becomes good enough to replace junior people, but can't quite do the big architecture and design things that senior engineers can do. In this world, a percentage of people will remain employed in software development. If this is the world you think will happen, you should gain skills as quickly as you can to always stay above the rising tide.

3. AI gets good, but remains a force multiplier for every worker, not a replacement. In economic terms, it's a "complement," not a "substitute." This is the world most people who doubt AI's job disruption fall into. They rightly say that in the past, Luddites were always wrong, that the progress we see does not eliminate jobs; it simply causes people to need to do different jobs. People in this belief system will quote things like the lump of labor fallacy.

I think there's no way to know which path is right now. Some of it depends on knowing yourself well enough to know what you will be differentially good at. In a world where change is happening much more rapidly than before, I think the person who survives best is not necessarily the person who guesses the future most accurately, but the person who changes most quickly to adapt as the environment changes. If you buy into this, it suggests not doing commitments of multiple years down certain paths unless you are sure that path is going to work out well. For instance, suppose you do get another college degree, taking three years to get a bachelor's in architecture or something you think will be safe from AI. You need to be pretty sure that that's the case because you've put a three-year dent on your ability to keep up with AI as you do all the schoolwork.

Wow. This is a tough and deep question.

We allow calculators in math classes now (they used to ban them "back in my day"), and yet we still teach basic arithmetic. I genuinely wonder why. (I don't mean that I don't agree -- I am literally simply asking, "What is the reason behind still teaching it?")

One could imagine a math class that was instead all about concepts. "Addition is combining two smaller things to make a bigger one." Etc. And then having calculators there on day one, like second grade. "Squares are the same number multiplied by themselves. It's like how you get the area of an actual square. Square roots are the opposite -- you go from knowing the area to wanting to know a side." Yes, it's true those kids would be completely screwed without a calculator — say, somehow stranded somewhere without technology and needing to add stuff, though I can't invent a reason right now. But is that rare instance worth spending years and years of class? Or would kids be able to solve much more interesting math problems, much more quickly, if we taught the concepts but not the mechanics?

It took a while for schools to stop teaching cursive. My kids were raised at a time when parents hotly debated this. Now it's accepted as a waste of time. "What if they have to write longhand?" On rare occasion, they're going to be much slower.

These days, we don't mind that most CS grads can't write a compiler themselves, or even look at assembly to decide whether there's a bug. But we accept that you'll typically always have a compiler when you need one, and that the world only needs a tiny number of people who can write you a new one.

I don't know if coding can/should go that way. Is there an AI-first, don't-worry-about-the-code way to teach programming to professionals in the field? Can we get to the point where AI almost serves the same function as a compiler, basically taking your higher-level language (English) and turning it into assembly? Once in a rare while, when it's wrong, perhaps only one engineer on your 50-person team can look at the code itself to fix the problem?

I'm brainstorming aloud here. Really not sure.