Yellow Toy Robot Manning AI Chat Stand

The Rapid Advancement of AI and Understanding Its Capabilities and Limitations

Digital By Sep 26, 2024 No Comments

Welcome to the first installment of our series based on my presentation at BiLT in Melbourne, titled “Top 10 Practical AI Workflows to Elevate Your Practice Today.” In this series, I plan to explore each topic covered in more detail, starting with a fundamental understanding of Artificial Intelligence (AI). Whether you’re a beginner, an expert, or just curious, this series aims to provide a clearer picture of how AI works and how it can be applied in the Architecture, Engineering, and Construction (AEC) industry. During the presentation, we had just 60 minutes to explore ten (well.. nine) exciting AI workflows, offering a glimpse of what’s possible. Over the next few weeks, we’ll take the time to unpack each one in detail, beginning with the basics of AI to set the stage for everything that follows.

Artificial Intelligence (AI) has come a long way in recent years. With the rise of large language models (LLMs) and computer vision technologies, AI-powered tools are becoming part of our everyday lives. This growth isn’t just happening in tech circles; professionals in fields like Architecture, Engineering, and Construction (AEC) are using AI to enhance their work. For example, architects might use AI to generate design options based on specific criteria, engineers could employ it to predict equipment failures before they happen, and construction managers might rely on AI for optimising project schedules.

But with all this talk about AI, Machine Learning (ML), and advanced models, it can get a bit confusing. What’s the difference between them? Think of AI as the big umbrella term—it’s about creating systems that can perform tasks that usually require human intelligence. Under that umbrella is Machine Learning, which is like teaching a computer to learn from data and improve over time without being explicitly programmed for every single task. Then we have specialized models like LLMs and vision models. These are advanced types of ML focused on specific areas like understanding and generating human language or interpreting visual information from images and videos.

So, how do these AI and ML models actually work? They learn by analysing large amounts of data to find patterns. Imagine reading thousands of books and then being able to predict how a story might unfold based on themes you’ve noticed. When you give these models new information, they make predictions or generate outputs based on what they’ve learned from all that data. But here’s the thing. These models don’t understand the content like we do. They can’t think original thoughts or grasp meaning in the way humans can. They mix and match existing information based on patterns they’ve seen before.

Strawberries

Let me give you an example. If you ask an AI language model, “How many Rs are in the word ‘strawberry’?” you might expect a quick and correct answer: two. But the model might get it wrong or provide a confusing response. That’s because of something called tokenization. The AI breaks down text into smaller pieces called tokens, which often represent chunks of words rather than individual letters. It works with these tokens using numbers, not the original letters. So, counting specific letters in a word isn’t something the AI naturally does well. This is a simple task for us but can trip up even advanced AI models.

Why does this matter? Knowing the limitations of AI helps us use it more effectively. While AI models are great at recognizing patterns and making predictions, they can stumble over tasks that are simple for humans. This doesn’t mean AI isn’t useful; it just means we need to be aware of where it excels and where it doesn’t. For instance, AI can quickly analyze large datasets to find trends that might take humans much longer to spot. But for tasks that require genuine understanding or common sense, human input remains incredibly important.

It’s important to keep in mind what these models can and can’t do. By understanding their capabilities and limitations, we can make the most of their strengths while compensating for their weaknesses. So, the next time you’re using an AI tool, remember that it’s a powerful assistant but not a replacement for human thought. Collaborating with AI can lead to better results, whether you’re designing a building, planning a project, or just trying to get an answer to a question.

No Comments

Leave a comment

Your email address will not be published. Required fields are marked *