Fundamentals of AI

Reading time 4 min

By Maurits Dijkgraaf

More and more, entrepreneurs and directors are rightly asking what AI exactly is and how it can strengthen their organization. In this article, I delve into the basics of AI. We hear about it constantly, various terms follow each other in rapid succession, but what does it all mean? I will answer questions such as: What is AI? How did it originate? What types of AI are there? What is machine learning? And what are large language models?

The Essence of AI

You might not expect it, but the roots of AI date back to the 1950s. At that time, the first computers were developed that could mimic simple human tasks, such as playing a game of chess. In the 1980s and 1990s, we saw growth in development, but real breakthroughs were lacking due to limited computing power and data.

With the advent of the internet and the explosive growth of data in the 21st century, AI received a new impetus. Data can be seen as the fuel of AI. Thanks to powerful computers and advanced algorithms, it is now possible to perform complex tasks, such as speech recognition and image analysis.

Instead of a computer only executing pre-programmed tasks, an AI system can recognize patterns, make decisions, and even make predictions based on data.

The Different Types of AI

When we talk about AI, it often remains vague which form of AI is meant. AI is a broad concept and can be divided into different categories:

Narrow AI (or Weak AI): Designed to perform one specific task. Examples include voice assistants like Siri or Alexa, which recognize your voice and execute commands.
General AI (or Strong AI): This type of AI can perform multiple tasks and has intelligence comparable to that of a human. Think of understanding emotions, creative thinking, or solving complex problems. This is currently still futuristic.
Superintelligence: This goes a step further than General AI and refers to AI that surpasses human intelligence in all respects. Although this is still theoretical, it does raise important ethical questions.

You may have noticed: we are currently in the era of Narrow AI. Despite the enormous hype surrounding AI, we are still only at the beginning of the possibilities. The systems we develop and use are specialized in specific tasks. They cannot function independently outside the parameters for which they are designed. General AI and Superintelligence are still pure future music.

Machine Learning

Another term you may often hear is Machine Learning. It is a subset of AI and revolves around the idea that systems can learn from data, recognize patterns, and make decisions with minimal human intervention. AI is thus powered by algorithms. There are three main forms:

Supervised Learning: Here, the system is trained with labeled data. Suppose you have photos of cats and dogs, and you label them as such. The system learns to recognize the difference and can correctly classify new photos.
Unsupervised Learning: The system receives unordered data without labels and must discover patterns itself. This is often used for customer segmentation, where you identify groups with similar behavior.
Reinforcement Learning: The system learns through trial-and-error and is rewarded for correct actions. This is widely applied in robotics and gaming.

The Power of Large Language Models

Another frequently used term in relation to AI is Large Language Models (LLMs). These are advanced AI models trained on enormous amounts of text. See it as a specific application of the previously discussed Machine Learning. LLMs understand context and can therefore generate human-like language that feels natural. A well-known example is GPT-4 from OpenAI, an AI model that anyone can use.

These models can:

Write texts: From blogs to articles and even poetry.
Answer questions: They can look up and summarize information.
Conduct conversations: Think of chatbots that provide customer service.
Make translations: They translate texts while preserving context and nuance.

I just briefly mentioned ChatGPT (or Chat) from OpenAI, but there are also other available applications that are quite successful based on LLMs. Below is a short overview, including ChatGPT from OpenAI for comparison. You see that these applications are similar but each chooses a slightly different approach.

ChatGPT (OpenAI): Used for text generation, but also customer service and writing creative content or brainstorming. Known for its natural language use and broad knowledge thanks to training on large datasets.
Google BERT (Bidirectional Encoder Representations from Transformers): Used for search queries and improving context understanding in search engines. Provides a better understanding of the meaning of words in sentences by analyzing context.
Claude (Anthropic): An assistant similar to ChatGPT and is used for Q&A, text generation, and more. Focuses on safe and ethical interactions with users.
Gemini (Google DeepMind): Applied in specific domains, such as coding assistants and research tools. Combines language understanding with advanced analysis of more complex tasks.
Copilot (GitHub): Assistant for software developers that generates code, detects errors, and accelerates development processes. Lowers the entry barrier for programming and increases the efficiency of experienced developers.

And Now? Don’t Start Too Quickly…

Now that we’ve discussed the basics of AI, you might already be thinking about potential applications for your business—custom software, new portals, anything is possible. Perhaps you can improve efficiency by automating tasks, make better decisions by analyzing data with AI and interacting with it in a chat, or significantly enhance a chatbot’s capabilities beyond pre-programmed questions and answers.

AI is no longer just a buzzword; it has become a practical tool that can genuinely help. However, be mindful that AI should not be a goal in itself for your organization. Carefully consider your ambitions and the objectives you want to achieve. From there, you can determine the right approach, with AI as a supporting tool—just as we would do with Rise.

Would you like to brainstorm about AI right away?