Everyone talks about AI. Everyone. My 6-year-old knows she can ask the computer to generate “a princess wearing a purple dress riding a unicorn” (true story).

But what exactly is AI?

As someone who works in tech and is submerged deep in AI, it took me quite some time to explain it to her. That got me thinking.

AI often sounds like something reserved for tech geeks. But let’s face it. It is already influencing our daily lives and its influence will only grow.

In this blog post, we’ll try to explore AI in a way that resonates with everyone.

Think of AI as…

Artificial intelligence is a set of algorithms. When those algorithms work on a task it can be compared to a sports fan (let’s call him John the fan) betting on a sports game. Both involve making informed predictions based on the analysis of available data and past experiences.

The AI processes vast quantities of historical data, discerning patterns, and evaluating variables to predict outcomes. Similarly, John the fan meticulously studies teams’ performances, individual player statistics, injuries, weather conditions, and even psychological factors before placing a bet.

Both the AI and John the fan leverage their respective forms of ‘intelligence’, experience, and available information to forecast future events. They refine their strategies over time, learning from past successes and failures. This process of continuous learning and adaptation is crucial in both fields.

Moreover, both AI and John the fan are subject to the complexities of real-world dynamics. For instance, just as unexpected events in a sports game — like a sudden injury or a surprising play — can derail the most well-researched bet, AI can be thrown off by anomalies or novel scenarios that weren’t present in its training data. These unexpected elements introduce a level of uncertainty and risk.

Despite these challenges, both AI and John the fan strive for accuracy. They adjust their models or strategies based on new information, always trying to minimize error and maximize predictive accuracy. However, the inherent unpredictability of real-world events means that neither AI nor sports betting can achieve perfect foresight. Thus, both fields require a balance of knowledge, intuition, and an acceptance of the limits of prediction.

This analogy underscores the complex interplay between data, prediction, and the ever-present possibility of unforeseen events that can influence outcomes in both artificial intelligence and sports betting.

Types of AI models

Diving deeper into the types of AI models, we can focus on a broad spectrum of categories including classification, segmentation, generative, regression, reinforcement learning, and anomaly detection models.

Each type serves a unique purpose and offers different methodologies for handling data and solving problems, much like how John the fan might choose specific strategies based on the bet’s type or the game he is betting on.

Classification Models are used in AI to categorize data into predefined classes. These models are critical in scenarios where decisions are distinctly categorical, such as distinguishing between spam and non-spam emails or identifying fraudulent bank transactions.

Regression Models are used for predicting a continuous outcome, such as stock prices or the amount of rainfall. These models are indispensable in fields that require the prediction of numerical values.

Anomaly Detection Models are designed to identify unusual patterns or outliers. These models are crucial in areas like fraud detection and network security, where recognizing deviations from normal patterns can prevent costly or dangerous situations.

Generative Models are designed to generate new data instances that resemble the training data. These models, such as Generative Adversarial Networks (GANs) or Variational Autoencoders (VAEs), are used for creating realistic synthetic images or generating new text instances, helping to understand the deep patterns within data.

Reinforcement Learning Models operate on a system of rewards and penalties, learning to perform tasks by trying to maximize cumulative reward. These models are used for complex decision-making tasks in robotics or strategic game play.

Each of these types of AI models brings its own set of tools and approaches, allowing AI developers to tailor their solutions to fit the complexities and nuances of specific challenges.

The choice among these models often depends on the specific requirements and constraints of the application, as each model type offers powerful capabilities for addressing distinct challenges in data analysis and prediction.

Last but not least

To be honest, I am still not sure this is a good enough or clear enough explanation. I am convinced that one of our missions is to help the construction industry learn more about AI and be informed about it so we will definitely keep trying.

Stay tuned for more blog posts about “what is AI”.

‍