Artificial intelligence (AI) is everywhere these days, and it’s not hard to see why. It’s revolutionizing industries, automating tasks, and making things possible that were once pure science fiction. But as I was reminded recently at one of our internal talks given by Fraser Gordon, a developer with a good deal of expertise in the area, while AI is incredible, it’s not actually intelligent—at least not in the way many people assume. It’s like that super eager coworker who’s always ready to help, but sometimes gives you an answer that’s… well, let’s just say creative.

At LiveCode, we’re big believers in AI—but we also know that to get the best out of it, you need more than just the technology. You need expertise. That’s the lesson I took away from a recent talk given by Fraser at our monthly staff get-together where we share insights and experiences. He broke down the intricacies of AI, how it works, and why human guidance is the secret ingredient that makes AI truly shine.

Fraser kicked things off by cutting through some of the hype surrounding AI. Contrary to what sci-fi movies might suggest, AI isn’t sitting there pondering the meaning of life. As he put it:

“AI is basically a very large pile of maths, but so much maths that it almost becomes magical.”

In other words, AI is number-crunching at an insane speed, but it doesn’t actually understand anything. When it processes text or data, it’s just running through a massive series of calculations to predict the most likely next word, sentence, or action based on patterns it’s seen before. AI is like throwing spaghetti at the wall to see what sticks—except in a much more sophisticated, math-heavy way. Sure, sometimes that spaghetti sticks perfectly, and the AI gets the right answer. But other times? It’s a miss. And without human expertise to guide it, AI can sometimes veer off into unexpected (and unhelpful) territory.

One of the most eye-opening parts of Fraser’s talk was when he dove into the definition of intelligence itself. He explained that intelligence is about more than just processing information—it’s about understanding, learning, and retaining knowledge. AI doesn’t quite hit the mark in any of these areas.

“Before we talk about what artificial intelligence is, we have to ask: what is intelligence? It’s about being able to perceive information, draw inferences, and learn from experience. AI doesn’t do any of that”

This really struck a chord with me because it’s easy to get caught up in AI’s dazzling capabilities and forget that it’s still very much a tool—not an independent thinker. AI lacks two key things that define real intelligence:

Understanding: AI processes data, but it doesn’t understand the meaning behind it. As Fraser said, “AI doesn’t understand written language in any meaningful way. It’s all probabilities and patterns, but there’s no comprehension of what’s actually being said.” AI can convincingly describe a raspberry but its never eaten one, and has no real understanding of the difference between, say, a raspberry and a jumbo jet.

Retention: AI doesn’t learn from its past interactions. Once a conversation or task is over, it’s like hitting reset. Fraser compared it to the goldfish myth. Contrary to popular belief, goldfish actually have pretty good memories. But AI doesn’t. “AI doesn’t retain information… It doesn’t remember anything at all. You start a conversation, you close a conversation, and you can’t ask it about the previous conversation. There is no retention of information”

So while AI can produce impressive results, it’s not “intelligent” in the human sense. It cannot reason itself into providing better answers. And that’s where human expertise comes in—guiding AI, providing context, and making sure it hits the right target.

“It works because it works. We don’t really understand how it works”

Fraser’s talk really highlighted something we’ve known for a while: while AI is powerful, it reaches its full potential when paired with human expertise. Here’s how we make that happen at LiveCode:

• Adding Context Where AI Can’t
  AI models don’t actually understand the world. They don’t have a mental model of how things work. As Fraser said:
  “AI doesn’t understand written language in any meaningful way.”
  It’s all just probabilities based on past data. That’s why, at LiveCode, we don’t just rely on AI to make decisions—we guide it with human insight. We help tailor AI solutions to your specific business needs, ensuring they deliver useful results, not just random guesses. Its all about the questions you ask, and the structure you provide for the answers. AI is great at picking from a menu of options. Its pretty terrible at, for example, designing user interface from scratch, because it has no spatial awareness, artistic ability, or understanding of UI norms. But given the guidance that usually, a “next” button is put at the bottom right of a screen, and a set of template options to pick from, AI can pick the right template and “design” a UI with “next” buttons in the right places.

• Fine-Tuning the Model
  AI might be great at processing data, but without proper training, it can easily go off course. AI isn’t truly intelligent; it’s more like a mimic, capable of outputting answers that look smart but often miss the mark.
  “It’s something that resembles intelligence but isn’t really intelligent at all,”
  Our job is to fine-tune those models, training them to behave the way you want, not how the data says they might. Like a very small child, AI needs told when it got it right, when it got it wrong, and how to change its output to be more right next time.

• Creating Seamless User Experiences
  AI might be powerful, but without proper integration, it’s like having a Ferrari engine in a go-kart—lots of power, but not exactly useful. In fact, I doubt if you’d like to ride in that go-kart. We make sure your AI fits seamlessly into your workflows, ensuring smooth and intuitive user experiences.
  “AI should enhance the user experience, not complicate it. Our job is to make sure it feels natural, like it was always there.”

• Ongoing Support (Because AI Doesn’t Run Itself)
  AI is not the “set it and forget it” tool some might think it is. This is a fast moving industry with improvements to the models shipping on a regular basis.
  “It’s very much an industry where things work because they work. We don’t always know why, but let’s keep going because its working in some magical fashion”
  While AI is powerful, it needs ongoing maintenance, tweaking, and support to keep delivering great results—and that’s where we come in. Our team is there every step of the way, checking the results and ensuring your AI stays sharp and relevant.

Fraser’s talk reinforced something we’ve always believed at LiveCode: AI is amazing, but without the right expertise, it’s just a really fast number cruncher. We’ve been through the AI journey ourselves, building and scaling our own products to users worldwide. We know firsthand what it takes to make AI solutions that don’t just sound smart—they are smart.

While AI can generate impressive results in the moment, it doesn’t learn from the past or apply context like a human would. That’s where human expertise comes in. We help your AI system become more than just a number-crunching machine—we make it a true asset to your business.

Ready to see how AI can transform your business, with the right guidance? Let’s talk. Together, we can create AI-powered solutions that don’t just throw spaghetti at the wall—they stick.