{"course":{"slug":"ai-actually-explained","title":"AI, Actually Explained","url":"https://addaly.com/learn/ai-actually-explained","level":"beginner","track":"start"},"lesson":{"slug":"the-family-tree","title":"Four rings, one inside the next","url":"https://addaly.com/learn/ai-actually-explained/the-family-tree","position":2,"of":92,"minutes":9,"answer":"AI, machine learning, deep learning and generative AI are four rings inside each other, and naming the ring tells you which question to ask: show me the rules, show me the data.","keyIdea":"AI, machine learning, deep learning and generative AI are four rings inside each other, and naming the ring tells you which question to ask: show me the rules, show me the data, or show me how you know the output is true."},"body":{"format":"markdown","text":"## The word covers too much\n\nWhen a shop sign, a job advert and a newspaper headline all say \"AI\", they are usually pointing at four different things. The four sit inside each other like rings. Once you can name which ring something is in, most of the confusion goes.\n\n**Artificial intelligence** is the outer ring. It is the oldest and vaguest of the four: any attempt to get a machine to do something that seems to need intelligence. A chess program from 1997 is in this ring. So is the routing that finds you a way round a traffic jam. Nothing about the ring says how the thing works.\n\n**Machine learning** sits inside it. Here, the behaviour was not written down by a person. It was derived from examples. Somebody collected a pile of data, ran a procedure over it, and the result decides what the system does. Your bank's fraud check is almost certainly in this ring.\n\n**Deep learning** sits inside machine learning. It is the subset that uses neural networks with many layers: the technique that took over from about 2012 onwards. Face unlock on your phone is deep learning. So is the speech recognition in a voice note.\n\n**Generative AI** sits inside deep learning. These are the models whose output is new content rather than a label or a number: text, images, audio, video. ChatGPT, Gemini, Midjourney, DeepSeek. This innermost ring is what most people now mean when they say AI, which is why the word has become so slippery: the newest and smallest ring has captured a name that belongs to the largest.\n\nLarge language models are a slice of that innermost ring. So a chatbot is a large language model, which is generative AI, which is deep learning, which is machine learning, which is artificial intelligence. Every one of those sentences is true, and each tells you something different.\n\n> Figure: Four rings, and the question each one licenses. Layers, top to bottom: Artificial intelligence (Anything that looks like it needs intelligence. Says nothing about how it works. Ask: show me the rules.); Machine learning (Behaviour derived from examples, not written down. Ask: what did you train it on, and how do you know it works?); Deep learning (Machine learning with many-layered networks. Took over from about 2012. Same questions, larger data.); Generative AI (Output is new content rather than a label. Add: how do you know the output is true? A paragraph has no right answer beside it.). Each ring sits inside the one above it. Naming the ring tells you what you are entitled to ask for: the rules, the data and the testing, or the evidence that the output is true.\n\n---\n\n## Why the ring matters\n\nBecause it tells you what question to ask.\n\nIf something is in the outer ring only (rules a person wrote) then you can ask to see the rules. Somebody can print them out. When a housing benefit system refuses your claim, and the logic is written rules, the refusal can be explained line by line.\n\nIf it is machine learning, the rules are not printable. The right question becomes: **what examples did you train it on, and how do you know it works?** Those two questions are the whole of the difference. Nobody can show you the reasoning, so they have to show you the data and the testing instead.\n\nIf it is generative, add a third question: **how do you know the output is true?** A label can be checked against the right label. A paragraph invented on the spot has no right answer sitting next to it.\n\n## The thing beginners get backwards\n\nMost people assume deep learning is better, and that anything serious must be a neural network by now. It is not so.\n\nFor data that lives in rows and columns (a spreadsheet of customers, a table of transactions, a sensor log) the method that usually wins is not deep learning at all. It is a family of models built from decision trees, with names like XGBoost, LightGBM and random forests. They train in minutes on an ordinary laptop, they need no graphics card, and on tabular data they routinely beat neural networks that cost a hundred times more to run. This has been tested repeatedly and it keeps coming out the same way.\n\nThis matters for you in a practical sense. If somebody tells you that predicting which customers will cancel requires a large language model, they are either selling something or have not looked. The free tools here are free and good: Python with `scikit-learn` and `xgboost`, or `LibreOffice Calc` for the first look at the data. The paid platforms (DataRobot, SageMaker, Vertex AI) are wrapping the same algorithms in a dashboard.\n\n## Where the rings are argued about\n\nBe warned that the rings are conventions, not laws. Three honest disagreements:\n\n- **Reinforcement learning** (learning by trial and reward, the method behind game-playing systems and part of how chatbots are tuned) does not sit neatly inside any of the four. Some diagrams place it beside machine learning, some inside it.\n- **A chess engine** like Stockfish scored positions with a hand-written evaluation from its first release in 2008 until 2020, when it added a small neural network. Since 2023 its evaluation has been entirely neural, though its search is still hand-written. It moved rings without changing its name.\n- **\"AI\" as a product label** is applied to plain database queries and `if` statements every day, because the word raises valuations and prices. There is no authority stopping this and no penalty for it.\n\nSo the rings are a thinking tool, not a certificate. Their use is that they replace one useless question with three useful ones.\n\n## The habit\n\nWhen you next meet a system described as AI, ask yourself which ring, out loud if you like. If you cannot tell, that is itself information: it means nobody has told you how the thing works, and you should treat every claim about it as unverified until they do."},"check":{"question":"A logistics firm wants to predict which of its 40,000 monthly deliveries will be late, using a spreadsheet of route, weather, driver and load data. A vendor proposes a large language model. What is the strongest technical objection?","options":[{"text":"Large language models cannot handle 40,000 rows, which is too much data for any model to process.","correct":false,"tempting":"Treats data volume as the obstacle; 40,000 rows is small, and volume is not what makes the proposal wrong."},{"text":"For data in rows and columns, tree-based models such as XGBoost usually predict better than neural networks, train in minutes on an ordinary laptop, and cost nothing.","correct":true},{"text":"Prediction of future events is outside the scope of artificial intelligence, which only classifies things that have already happened.","correct":false,"tempting":"Invents a boundary that does not exist; prediction of future events is the main commercial use of machine learning."},{"text":"A language model would work, but the firm should first get a larger dataset, because deep learning always needs millions of examples.","correct":false,"tempting":"Accepts the vendor's framing and treats the shortfall as data size, when the mismatch is between the method and the shape of the problem."}]},"prev":{"slug":"the-word-ai","title":"What the word actually names","url":"https://addaly.com/learn/ai-actually-explained/the-word-ai"},"next":{"slug":"written-versus-trained","title":"Written by a person, or grown from examples","url":"https://addaly.com/learn/ai-actually-explained/written-versus-trained"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Four rings, one inside the next', in AI, Actually Explained. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-actually-explained/the-family-tree (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Four rings, one inside the next","courseTitle":"AI, Actually Explained","module":{"index":1,"slug":"what-the-word-names","title":"What this word actually names"},"answer":null,"keyIdea":"AI, machine learning, deep learning and generative AI are four rings inside each other, and naming the ring tells you which question to ask: show me the rules, show me the data, or show me how you know the output is true.","hero":{"kind":"figure","figure":{"t":"figure","label":"Figure","figure":{"type":"stack","title":"Four rings, and the question each one licenses","layers":[{"label":"Artificial intelligence","detail":"Anything that looks like it needs intelligence. Says nothing about how it works. Ask: show me the rules."},{"label":"Machine learning","detail":"Behaviour derived from examples, not written down. Ask: what did you train it on, and how do you know it works?"},{"label":"Deep learning","detail":"Machine learning with many-layered networks. Took over from about 2012. Same questions, larger data."},{"label":"Generative AI","detail":"Output is new content rather than a label. Add: how do you know the output is true? A paragraph has no right answer beside it."}],"caption":"Each ring sits inside the one above it. Naming the ring tells you what you are entitled to ask for: the rules, the data and the testing, or the evidence that the output is true."},"alt":"Figure: Four rings, and the question each one licenses. Layers, top to bottom: Artificial intelligence (Anything that looks like it needs intelligence. Says nothing about how it works. Ask: show me the rules.); Machine learning (Behaviour derived from examples, not written down. Ask: what did you train it on, and how do you know it works?); Deep learning (Machine learning with many-layered networks. Took over from about 2012. Same questions, larger data.); Generative AI (Output is new content rather than a label. Add: how do you know the output is true? A paragraph has no right answer beside it.). Each ring sits inside the one above it. Naming the ring tells you what you are entitled to ask for: the rules, the data and the testing, or the evidence that the output is true."}},"blocks":[{"t":"h2","id":"the-word-covers-too-much","runs":[{"text":"The word covers too much"}]},{"t":"p","runs":[{"text":"When a shop sign, a job advert and a newspaper headline all say “AI”, they are usually pointing at four different things. The four sit inside each other like rings. Once you can name which ring something is in, most of the confusion goes."}]},{"t":"p","runs":[{"text":"Artificial intelligence","b":true},{"text":" is the outer ring. It is the oldest and vaguest of the four: any attempt to get a machine to do something that seems to need intelligence. A chess program from 1997 is in this ring. So is the routing that finds you a way round a traffic jam. Nothing about the ring says how the thing works."}]},{"t":"p","runs":[{"text":"Machine learning","b":true},{"text":" sits inside it. Here, the behaviour was not written down by a person. It was derived from examples. Somebody collected a pile of data, ran a procedure over it, and the result decides what the system does. Your bank’s fraud check is almost certainly in this ring."}]},{"t":"p","runs":[{"text":"Deep learning","b":true},{"text":" sits inside machine learning. It is the subset that uses neural networks with many layers: the technique that took over from about 2012 onwards. Face unlock on your phone is deep learning. So is the speech recognition in a voice note."}]},{"t":"p","runs":[{"text":"Generative AI","b":true},{"text":" sits inside deep learning. These are the models whose output is new content rather than a label or a number: text, images, audio, video. ChatGPT, Gemini, Midjourney, DeepSeek. This innermost ring is what most people now mean when they say AI, which is why the word has become so slippery: the newest and smallest ring has captured a name that belongs to the largest."}]},{"t":"p","runs":[{"text":"Large language models are a slice of that innermost ring. So a chatbot is a large language model, which is generative AI, which is deep learning, which is machine learning, which is artificial intelligence. Every one of those sentences is true, and each tells you something different."}]},{"t":"rule"},{"t":"h2","id":"why-the-ring-matters","runs":[{"text":"Why the ring matters"}]},{"t":"p","runs":[{"text":"Because it tells you what question to ask."}]},{"t":"p","runs":[{"text":"If something is in the outer ring only (rules a person wrote) then you can ask to see the rules. Somebody can print them out. When a housing benefit system refuses your claim, and the logic is written rules, the refusal can be explained line by line."}]},{"t":"p","runs":[{"text":"If it is machine learning, the rules are not printable. The right question becomes: "},{"text":"what examples did you train it on, and how do you know it works?","b":true},{"text":" Those two questions are the whole of the difference. Nobody can show you the reasoning, so they have to show you the data and the testing instead."}]},{"t":"p","runs":[{"text":"If it is generative, add a third question: "},{"text":"how do you know the output is true?","b":true},{"text":" A label can be checked against the right label. A paragraph invented on the spot has no right answer sitting next to it."}]},{"t":"h2","id":"the-thing-beginners-get-backwards","runs":[{"text":"The thing beginners get backwards"}]},{"t":"p","runs":[{"text":"Most people assume deep learning is better, and that anything serious must be a neural network by now. It is not so."}]},{"t":"p","runs":[{"text":"For data that lives in rows and columns (a spreadsheet of customers, a table of transactions, a sensor log) the method that usually wins is not deep learning at all. It is a family of models built from decision trees, with names like XGBoost, LightGBM and random forests. They train in minutes on an ordinary laptop, they need no graphics card, and on tabular data they routinely beat neural networks that cost a hundred times more to run. This has been tested repeatedly and it keeps coming out the same way."}]},{"t":"p","runs":[{"text":"This matters for you in a practical sense. If somebody tells you that predicting which customers will cancel requires a large language model, they are either selling something or have not looked. The free tools here are free and good: Python with "},{"text":"scikit-learn","code":true},{"text":" and "},{"text":"xgboost","code":true},{"text":", or "},{"text":"LibreOffice Calc","code":true},{"text":" for the first look at the data. The paid platforms (DataRobot, SageMaker, Vertex AI) are wrapping the same algorithms in a dashboard."}]},{"t":"h2","id":"where-the-rings-are-argued-about","runs":[{"text":"Where the rings are argued about"}]},{"t":"p","runs":[{"text":"Be warned that the rings are conventions, not laws. Three honest disagreements:"}]},{"t":"ul","items":[[{"text":"Reinforcement learning","b":true},{"text":" (learning by trial and reward, the method behind game-playing systems and part of how chatbots are tuned) does not sit neatly inside any of the four. Some diagrams place it beside machine learning, some inside it."}],[{"text":"A chess engine","b":true},{"text":" like Stockfish scored positions with a hand-written evaluation from its first release in 2008 until 2020, when it added a small neural network. Since 2023 its evaluation has been entirely neural, though its search is still hand-written. It moved rings without changing its name."}],[{"text":"”AI” as a product label","b":true},{"text":" is applied to plain database queries and "},{"text":"if","code":true},{"text":" statements every day, because the word raises valuations and prices. There is no authority stopping this and no penalty for it."}]]},{"t":"p","runs":[{"text":"So the rings are a thinking tool, not a certificate. Their use is that they replace one useless question with three useful ones."}]},{"t":"h2","id":"the-habit","runs":[{"text":"The habit"}]},{"t":"p","runs":[{"text":"When you next meet a system described as AI, ask yourself which ring, out loud if you like. If you cannot tell, that is itself information: it means nobody has told you how the thing works, and you should treat every claim about it as unverified until they do."}]}],"check":{"question":[{"text":"A logistics firm wants to predict which of its 40,000 monthly deliveries will be late, using a spreadsheet of route, weather, driver and load data. A vendor proposes a large language model. What is the strongest technical objection?"}],"options":[{"text":[{"text":"Large language models cannot handle 40,000 rows, which is too much data for any model to process."}],"correct":false,"tempting":[{"text":"Treats data volume as the obstacle; 40,000 rows is small, and volume is not what makes the proposal wrong."}]},{"text":[{"text":"For data in rows and columns, tree-based models such as XGBoost usually predict better than neural networks, train in minutes on an ordinary laptop, and cost nothing."}],"correct":true,"tempting":null},{"text":[{"text":"Prediction of future events is outside the scope of artificial intelligence, which only classifies things that have already happened."}],"correct":false,"tempting":[{"text":"Invents a boundary that does not exist; prediction of future events is the main commercial use of machine learning."}]},{"text":[{"text":"A language model would work, but the firm should first get a larger dataset, because deep learning always needs millions of examples."}],"correct":false,"tempting":[{"text":"Accepts the vendor’s framing and treats the shortfall as data size, when the mismatch is between the method and the shape of the problem."}]}],"order":[2,0,3,1],"why":null},"prev":{"slug":"the-word-ai","title":"What the word actually names","minutes":6},"next":{"slug":"written-versus-trained","title":"Written by a person, or grown from examples","minutes":7}}}