Can AI Really Write Code? Let's Find Out

If someone tells you that AI can write all the code you need, they're not telling the truth.

This isn't just a guess. It's not an exaggeration. And it's not meant to scare you.

We now know enough about how AI works, especially the new kind called GenAI, to say this with certainty. I'm not just talking about what we've learned in the last two years, but what we've learned over the last 20 years. I was there when it all started. I know what I'm talking about.

Many of you already know this too. But your job isn't done yet. You need to help explain why using AI to write real code for apps that customers will use is like opening a restaurant with nothing but a bunch of pretty recipe cards.

They might look nice, but you can't eat paper.

To help everyone understand, let me ask you a question:

Q: How would you know if AI wrote this article?

A: Because it would suck.

Maybe AI could copy my style and use some of my favorite writing tricks, but we're not sure if it can really write like me for more than a sentence or two.

Banana. Take that, AI!

I'm 100% sure AI can't take my years of experience in the topics I write about - topics that change all the time in the tech and business world - and use my sarcastic words to share useful ideas with the most people who would find them helpful.

That's structure. It's the foundation. It's boring. But it's the only thing that holds these bits of writing together.

Look, if you want to write about a technical or entrepreneurial topic, you either need to a) spend a lifetime doggedly nerding down those paths with real-world, real-life stakes and consequences, or b) read a bunch of articles written by people who have done just that and then summarize those articles as best you can without understanding half of what those people are actually talking about.

Which one sounds more like AI, a) or b)?

Now let's talk about how this relates to code, because I hope you can already see the connection.

Real coders know this.

The threat that AI presents to your average software developer is not new.

Raise your hand if you've ever used GitHub or Stack Overflow or any other kind of example code or library or whatever to help you get started on the foundational solution to the business problem that your code needs to solve.

Now, put your hand down if you've never once had to spend hours, sometimes days, tweaking and modifying that sample code a million times over to make it work like you need it to work to solve your unique problem.

OK. All of you who put your hands down. Get out of the room. Seriously. Go. We can't have a serious discussion about this.

Cheap, flawed code that's easy to break has been a threat to software developers since they first let us kids play with Basic - not to mention the threat of any tech solution that ends with "-shoring".

The AI threat just seems existential because of the constant repetition of a few exaggerated truths. That it's "free," that it's "original," and that it "works."

Here's why that's going to be a race to failure. Position yourself.

That's the most judgmental, sarcastic header I've ever written. But there's a reason why every word is in quotes. Because this is how the lie spreads.

Yes, what we're calling AI today does a decent job of putting code together in a way that runs. I'm not even going to talk about the big difference between GenAI and real AI or why code is more than just syntax.

But I will point out that - even beyond those issues - we're not at anything I'd call usable yet. Some worrying words from an IEEE study follow:

[ChatGPT has] a success rate ranging from as low as 0.66 percent to as high as 89 percent -- depending on how hard the task is, what programming language it's using, and a few other things.

I'll let you decide how "hard," "coding language," and "other stuff" affect how well it works. Sorry for the quotes again. If it helps, I almost hurt my finger because I was waving my hands so much while reading that thing.

A study conclusion (my italics): "ChatGPT hasn't seen new problems and answers yet. It can't think critically like a human and can only handle problems it's seen before."

Just like I said about why AI-written articles are bad, if you're trying to solve new problems by making new answers, AI has no experience with this.

Okay, all you "ChatGPT-4-is-The-One" fans can argue with me now. But AI doesn't just have problems with the code writing.

Code alone is useless.

Every coder reading this just said, "Yep."

Besides all the problems AI has when it makes code out of "nothing" (or to use the tech term, "other people's code"), bigger issues show up when we try to get that code to customers.

Code without design, how it looks, how it works, what it needs to do, and what the business needs is just a classroom exercise that doesn't matter. The problem AI has with any of those "other important" success factors is that none of them are simple yes or no. None.

For example, Figma had to stop its AI design feature for a while when people said its AI was just copying someone else's design.

"Just tell us what you need, and the feature will give you a first try," is how the company explained it when they started the feature.

I can do that without AI. I can do that by copying and pasting. Figma blamed bad testing. Which one sounds more true?

But not elegance.

If your code doesn't have a smooth connection from the boring structure work to the design and how it works for customers, you can still call it "code" if you want, but it will be as valuable as an AI-made avatar reading AI-made content over AI-made pictures.

Have you ever seen that? It'll hurt your soul.

There's a right way to do things and there's a way to do things well, and I know sometimes you can't do both. But this is 30 years of tech history happening again, and the tech people need to start teaching history or we'll keep having to repeat it.

So I'd ask my coder friends to raise your hand if you've ever had to come in and fix someone's badly made, often broken, messy, and not smooth code.

Okay. Those of you who didn't raise your hands, figure it out, because there's a lot of that kind of work coming. And anyone who has ever had to fix bad code can tell you it takes much longer to do that than it would have taken to just code it well from the start.

I'm going to talk more about humans using AI the right way to solve problems for other humans. If you're interested in that kind of thing, now would be a good time to join my email list at joeprocopio.com.

CREDIT: Thanks to Joe Procopio, Founder, TeachingStartup.com