This article is for you \ud83d\ude09 <\/p>\n
The objective here is clarity<\/strong> \u2014 not simplification through approximation, but accurate explanation in plain language. No technical background is required for the rest of the article.<\/p>\n The key idea: the Matryoshka doll<\/p>\n A useful way to understand the relationship between Artificial Intelligence, Machine Learning, Deep Learning, and Generative AI is to imagine Matryoshka dolls<\/strong>.<\/p>\n Let\u2019s open them one by one.<\/p>\n Artificial Intelligence: the outer shell<\/p>\n Artificial Intelligence (AI)<\/strong> is the broadest definition. 1. Make decisions. Example: A navigation system choosing the fastest route based on real-time traffic conditions. Rule-based AI: intelligence written by humans<\/p>\n In the early decades of AI, particularly in the 1970s and 1980s, systems were primarily rule-based<\/strong>. What I mean is that humans explicitly wrote the logic. The computer did not learn \u2014 it executed predefined instructions.<\/p>\n This was considered Artificial Intelligence because human reasoning was encoded and executed entirely by a machine<\/strong>.<\/p>\n Why rule-based AI was limited<\/p>\n Rule-based systems work well only in controlled environments. The system cannot adapt unless a human rewrites the rules. Machine Learning: letting data speak<\/p>\n Machine Learning (ML)<\/strong> is a subset of Artificial Intelligence. Instead of telling the computer what the rules are, we let the system learn them directly from examples.<\/p>\n -> Let\u2019s return to the house price example. Instead of writing rules, we collect data:<\/p>\n Thousands, sometimes millions, of past examples.<\/p>\n This data is provided as training data to a machine learning model.<\/p>\n But what does \u201ctraining a model\u201d using data actually mean?<\/p>\n Training is not a black box. We begin by choosing a mathematical model<\/strong> \u2014 essentially an equation \u2014 that could describe the relationship between inputs (surface, location, etc.) and output (price).<\/p>\n We do not test one equation. We test many (We call them models). The model adjusts its parameters by comparing prices with real prices across many examples.<\/p>\n No human could manually analyze hundreds of thousands of houses at once. A machine can.<\/p>\n How do we know a model works?<\/p>\n Before adopting a model \u2014 that is, the equation that best represents the phenomenon we are studying \u2014 we evaluate it. If performance is good on unseen data, the model is useful. Machine learning excels at tasks humans struggle with:<\/p>\n The limits of traditional machine learning<\/p>\n Nevertheless, traditional Machine Learning has important limitations. It works very well with structured data<\/strong>:<\/p>\n However, it struggles with types of data that humans handle naturally, such as:<\/p>\n The reason for this limitation is fundamental -> \u201ccomputers only understand numbers<\/strong>\u201c<\/p>\n Computers do not understand images, sounds, or words the way humans do.<\/p>\n They only understand numbers<\/strong>.<\/p>\n When working with images, text, or audio, these data must first be transformed into numerical representations.<\/p>\n For example, an image is converted into a matrix of numbers<\/strong>, where each value corresponds to pixel information such as color intensity. Only after this conversion can a machine learning model process the data.<\/p>\n This transformation step is mandatory.<\/p>\n Feature extraction: the traditional approach<\/p>\n Before the rise of deep learning, this transformation relied heavily on manual feature engineering<\/strong>.<\/p>\n Engineers had to decide in advance which characteristics might be useful:<\/p>\n This process, known as feature extraction<\/strong>, was:<\/p>\n Small changes in the data often required redesigning the features from scratch.<\/p>\n Why deep learning was needed<\/p>\n The limitations of manual feature extraction in complex settings were a key motivation for the development of Deep Learning. (I\u2019m not covering the more technical motivations in this article. My goal is to give you a clear understanding of the big picture).<\/p>\n Deep Learning does not eliminate the need for numerical data. Rather than relying on hand-crafted features designed by humans, deep learning models learn useful representations directly from raw data<\/strong>.<\/p>\n This marks a structural shift.<\/p>\n Deep Learning: the structural shift<\/p>\n Deep Learning still works as Machine Learning. The learning process is the same: What changes is what we call the architecture of the model<\/strong>. Layers as progressive representations<\/p>\n Each layer in a deep learning model applies a mathematical transformation to its input and passes the result to the next layer.<\/p>\n These layers can be understood as progressive representations<\/strong> of the data.<\/p>\n In the case of image recognition:<\/p>\n The model does not \u201csee\u201d images the way humans do. Instead of being told explicitly which features to use, the model learns them directly from the data.<\/p>\n This ability to automatically learn representations<\/strong> is what makes deep learning effective for complex, unstructured data (see the representation above).<\/p>\n And once this level of understanding is reached, an important shift becomes possible.<\/p>\n Up to this point, deep learning models have mainly been used to analyze existing data<\/strong>.<\/p>\n They are trained to:<\/p>\n In short, they help answer the question: What is this?<\/p>\n But learning rich representations of data naturally raises a new question:<\/p>\n If a model has learned how data is structured, could it also produce new data<\/strong> that follows the same structure?<\/p>\n<\/blockquote>\n This question is the foundation of Generative AI<\/strong>.<\/p>\n Generative AI: from analysis to creation<\/p>\n Generative AI does not replace deep learning. It builds directly on top of it.<\/p>\n The same deep neural networks that learned to recognize patterns can now be trained with a different objective: generation<\/strong>.<\/p>\n Instead of focusing only on classification or prediction, generative models learn how data is produced, step by step.<\/p>\n As a result, they are able to create new content that is coherent and realistic.<\/p>\n A concrete example<\/p>\n Consider the prompt:<\/p>\n \u201cDescribe a luxury apartment in Paris.\u201d<\/p>\n<\/blockquote>\n The model does not retrieve an existing description.<\/p>\n Instead:<\/p>\n Each prediction depends on:<\/p>\n The final text is new \u2014 it has never existed before \u2014 yet it feels natural because it follows the same structure as similar texts seen during training.<\/p>\n The same principle across data types<\/p>\n This mechanism is not limited to text. The same generative principle applies to:<\/p>\n This is why these models are often called foundation models<\/strong>: a single trained model can be adapted to many different tasks.<\/p>\n Why Generative AI feels different today<\/p>\n Artificial Intelligence, Machine Learning, and Deep Learning have existed for many years.<\/p>\n What makes Generative AI feel like a turning point is not only improved performance, but how humans interact with AI<\/strong>.<\/p>\n In the past, working with advanced AI required:<\/p>\n Today, interaction happens primarily through:<\/p>\n Users no longer need to specify how to do something. This shift dramatically reduces the barrier to entry and allows AI to integrate directly into everyday workflows across a wide range of professions.<\/p>\n Putting everything together<\/p>\n These concepts are not competing technologies. They form a coherent progression:<\/p>\n Seen this way, Generative AI is not a sudden break from the past. Once this structure is clear, AI terminology stops being confusing and becomes a coherent story.<\/p>\n But, Have we finished? Almost.<\/p>\n At this point, we\u2019ve covered the core AI ecosystem<\/strong>: artificial intelligence, machine learning, deep learning, and generative AI \u2014 and how they naturally build on one another.<\/p>\n If you are reading this article, there is a good chance you already use tools like ChatGPT in your daily life. I won\u2019t go much deeper here \u2014 this deserves an article of its own.<\/p>\n However, there is one important final idea worth remembering.<\/p>\n Earlier, we said that Generative AI is a continuation of Deep Learning<\/strong>, specialized in learning patterns well enough to generate new data that follows those same patterns.<\/p>\n That is true \u2014 but when it comes to language<\/strong>, the patterns involved are far more complex.<\/p>\n Human language is not just a sequence of words. It is structured by grammar, syntax, semantics, context, and long-range dependencies. Capturing these relationships required a major evolution in deep learning architectures.<\/p>\n From Deep Learning to Large Language Models<\/p>\n To handle language at this level of complexity, new deep learning architectures emerged. These models are known as Large Language Models (LLMs)<\/strong>.<\/p>\n Instead of trying to understand the full meaning of a sentence all at once, LLMs learn language in a very particular way:<\/p>\n They learn to predict the next word (or token) given everything that comes before it<\/strong>.<\/p>\n This might sound simple, but when trained on massive amounts of text, this objective forces the model to internalize:<\/p>\n By repeating this process billions of times, the model learns an implicit representation of how language works.<\/p>\n From these Large Language Models, conversational systems such as ChatGPT<\/strong> are built \u2014 combining language generation with instruction-following, dialogue, and alignment techniques.<\/p>\n The illustration above shows this idea visually: generation happens one word at a time<\/strong>, each step conditioned on what was generated before.<\/p>\n The final big picture<\/p>\n Nothing you see today came out of nowhere.<\/p>\n ChatGPT is not a separate technology. It is the visible result of a long progression:<\/p>\n I hope this article was helpful. And now, you\u2019re no longer lost in tech conversations \u2014 even at your end-of-year family gatherings \ud83d\ude42<\/p>\n If you enjoyed this article, feel free to follow me on LinkedIn for more honest insights about AI, Data Science, and careers.<\/p>\n \ud83d\udc49\u00a0LinkedIn:<\/strong>\u00a0Sabrine Bendimerad<\/a>![\"\"](\"https:\/\/www.europesays.com\/ie\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-01-at-16.04.40-1024x781.png\")
Figure 1 \u2014 AI Ecosystem (created by the author using Excalidraw)<\/p>\n\n
At its core, AI refers to systems designed to perform tasks that typically require human intelligence. In practice, AI includes systems that can:<\/p>\n
2. Draw conclusions. Example: A system deciding whether to approve or reject a loan application based on multiple factors.
3. Recognize patterns. Example: Detecting fraudulent credit card transactions by identifying unusual spending behavior.
4. Predict outcomes. Example: Estimating future energy consumption or product demand.<\/p>\n\n
Real-world conditions are not controlled. If we are still with our real estate example.<\/p>\n\n
This limitation led to the next layer.<\/p>\n
The key shift is simple but profound:<\/p>\n\n
![\"\"](\"https:\/\/www.europesays.com\/ie\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-01-at-15.00.03-1024x752.png\")
Table 1 \u2014 Example of Data (created by the author using Excalidraw)<\/p>\n![\"\"](\"https:\/\/www.europesays.com\/ie\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-02-at-10.12.34-1024x250.png\")
Figure 2 \u2014 Machine Learning Pipeline (created by the author using Excalidraw)<\/p>\n
A very simplified example might look like:
price = 2 \u00d7 surface + 3 \u00d7 location<\/p>\n
Part of the data is intentionally hidden. This is known as test data<\/strong>.
The model:<\/p>\n\n
![\"\"](\"https:\/\/www.europesays.com\/ie\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-02-at-10.11.34-1024x485.png\")
Figure 3 \u2014 Train\/Test Data (created by the author using Excalidraw)<\/p>\n
If not, it is discarded and another model is tried.
This evaluation step is essential.<\/p>\n\n
\n
-> disease risk prediction,
-> analysis of medical images.<\/li>\n
-> predicting equipment failures,
-> optimizing production processes.<\/li>\n
-> recommendation systems,
-> fraud detection.<\/li>\n<\/ol>\n\n
![\"\"](\"https:\/\/www.europesays.com\/ie\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-02-at-10.19.04-1024x281.png\")
Figure 4 \u2014 Converting an Image to Matrices (created by the author using Excalidraw)<\/p>\n\n
\n
Instead, it changes how features are obtained<\/strong>.<\/p>\n
-> data,
-> training,
-> evaluation.<\/p>\n
Deep learning relies on neural networks with many layers<\/strong>.<\/p>\n![\"\"](\"https:\/\/www.europesays.com\/ie\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-01-at-15.38.07-1024x400.png\")
Figure 5 \u2014 Deep Learning Pipeline (created by the author using Excalidraw)<\/p>\n\n
It learns a hierarchy of numerical representations that make accurate predictions possible.<\/p>\n\n
\n
![\"\"](\"https:\/\/www.europesays.com\/ie\/wp-content\/uploads\/2025\/12\/image-8-1024x547.png\")
Figure 6 \u2014 GenAI Pipeline (created by the author using Gemini3)<\/p>\n\n
\n
\n
\n
\n
\n
They can simply describe what they want.<\/p>\n\n
It is the natural continuation of everything that came before.<\/p>\n![\"\"](\"https:\/\/www.europesays.com\/ie\/wp-content\/uploads\/2025\/12\/Screenshot-2025-12-02-at-10.53.06-1024x935.png\")
Figure 6 \u2014 The Complete AI Ecosystem in 2025 (created by the author using Excalidraw) <\/p>\n\n
\n
\ud83d\udc49\u00a0Medium:<\/strong>\u00a0https:\/\/medium.com\/@sabrine.bendimerad1<\/a><\/p>\n","protected":false},"excerpt":{"rendered":". Machine Learning and Deep Learning are mentioned just as often. And now, Generative AI seems to dominate…\n","protected":false},"author":2,"featured_media":220667,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[261],"tags":[291,289,290,6668,612,18,5568,19,17,14338,610,82],"class_list":{"0":"post-220666","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-artificial-intelligence","8":"tag-ai","9":"tag-artificial-intelligence","10":"tag-artificialintelligence","11":"tag-data-science","12":"tag-deep-learning","13":"tag-eire","14":"tag-generative-ai","15":"tag-ie","16":"tag-ireland","17":"tag-llm","18":"tag-machine-learning","19":"tag-technology"},"share_on_mastodon":{"url":"https:\/\/pubeurope.com\/@ie\/115680054671470237","error":""},"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts\/220666","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/comments?post=220666"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts\/220666\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/media\/220667"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/media?parent=220666"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/categories?post=220666"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/tags?post=220666"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}