Ai

“I just ran it through ChatGPT,” my teammate said, “and this came out.” An online presentation followed. There were no questions.

In the Andrej Karpathy video I shared earlier this quote came up: “You can outsource your thinking, but not your understanding.” I think that’s too optimistic. Human nature is just to optimize for laziness.

AI makes mistakes all the time, or even lies to you, or makes up facts, or is based on outdated information. It’s constantly in the news, and everyone knows it. AI cites an article about a completely fabricated disease. Lawyers showing up with nonsense sources. And yet everyone keeps asking it everything, uncritically, and believing the answer. Because it sounds like a human.

For the insiders Andrej Karpathy needs no introduction, for the rest: Karpathy is among the very best in AI worldwide. He co-founded OpenAI and led AI and AutoPilot Vision at Tesla. He’s also a master at explaining things, and makes really fun and accessible videos on YouTube with titles like “Let’s build GPT: from scratch, in code, spelled out”. Lately he’s mostly focused on everything around LLMs, and he coined the term ‘Vibe coding’, which means ‘programming by just prompting’. You tell an LLM like ChatGPT or Claude what you want, and the LLM makes it for you. A great way for anyone, also non-programmers, to make software for the specific things you want to solve. Low barrier, online.

In February 1996, I came up with a way to measure and report website traffic remotely. I put an image on my ‘HTML Corner’ website, served from my own server, named it Nedstat (the directory needed a name), and built the first version. When the page loaded, a file was updated on the server. Clicking the image triggered a Perl script that reported: last 10 visitors (due to a bug it was 11, don’t ask) visitors per day, per hour, per country. Real time. Simple. But this idea didn’t exist yet. I thought immediately: this could be something big.

I emailed the ‘Surprising Site of the Day’, a Dutch cool-site-of-the-day equivalent, run by Hans Veldhuizen. We already knew each other; I built a lot of sites and had been featured before. He wrote back: HTML Corner is nice, tomorrow you’re SSotD, but what’s on it, that Nedstat, that beats everything. We need to talk.

I’ve been spending the past few months coding with AI. Here’s what I’ve learned so far.

Design before you delegate

Your own job is to build a smart architecture and data model. Use your unique creative thinking, and your own experience. For example, I never want to deal with timezones, so everything is always UTC until it’s displayed. That’s a decision the AI won’t make for you.

But you can’t plan everything upfront. You will iterate. Sometimes your idea just doesn’t work. This means you need to refactor all the time. It’s ok. Make many small refactors. Don’t wait until big ones are needed.

Experts speak. A Robotics professor and a researcher in Applied Mathematics from Johns Hopkins University discuss how AI can aid in the fight against climate change

When you read the title, you might wonder: How can something as virtual and abstract as AI assist in combating something as vast and tangible as climate change? Fortunately, the speakers have developed a clever framework that makes this understandable and logical by breaking it down into three clear areas: understanding the problem, predicting the direction, and assisting in the required transformation.

1. Understand the Problem

What do we know for certain? CO2 levels are rising, and the greenhouse effect is real. However, data collection is not very detailed. In the ocean, AI aids in operating autonomous underwater robots equipped with sensors. Remote control is impractical because radio waves do not work underwater, making AI a useful tool for enabling independent navigation.