{
  "video_id": "phOhGqpXss4",
  "channel_slug": "ibmtechnology",
  "channel_handle": "IBM Technology",
  "title": "Predictive vs Generative AI: How They Work and When to Use Each",
  "duration_seconds": 719,
  "url": "https://www.youtube.com/watch?v=phOhGqpXss4",
  "upload_date": "20260511",
  "transcript": "Predictive AI, generative AI, sometimes people use these terms interchangeably.\nThey shouldn't because these are fundamentally different tools that answer different questions.\nSo predictive AI, that asks the question specifically, what will happen?\nWhat's going to happen next?\nIt looks at historical data, and it forecasts an outcome.\nIs this transaction fraudulent or how many units will we sell next quarter?\nSo predictive AI is about making predictions about specific measurable things.\nGenerative AI, that answers a pretty different question, which is what could this look like, which is fundamentally different to what will happen.\nGenerative AI creates new content that resembles its training data.\nSo write me an email or help me debug this code.\nSo unlike predictive AI, it's not mapping to a specific label or value.\nInstead it's producing or you know generating something that didn't exist before.\nNow what do these things actually output?\nWell, predictive AI, that gives you perhaps a number, or it gives you a category, like put this thing in the right box.\nIt gives you a probability, something you can measure and verify, like there's a 87% chance this customer will cancel their subscription.\nThat's something you can check on later on.\nIt can be right or it can be wrong.\nGenerative AI gives you content.\nSo the content that it gives you, might be some text like some sentences or a paragraph.\nIt could be an AI-generated image.\nIt could a piece of code.\nAnd in many cases, there isn't a single correct answer.\nIf we ask a generative model to write a product description, well, there are thousands of valid ways to do that.\nIt's subjective.\nAnd what about the data that these systems consume?\nWell, predictive AI largely consumes structured data.\nThis is rows and columns or database tables or sensor readings, stuff with nice, clean labels.\nGenerative AI, well, that largely consumes unstructured data.\nSo this is kind of the messy stuff, billions of words scraped from the internet or pixels in photos or random code snippets.\nGenerative AI takes this kind of chaos and finds patterns in it, which it can use to generate something new.\nSo a question for you.\nWhere do large language models fit into this picture?\nGenerative, right?\nWell, right at this very moment, there's somebody tapping away in the comments section of this video saying large language models predict the next token,\nso LLMs are both generative and predictive.\nWell, first, thank you for commenting.\nPlease also remember to like and subscribe.\nAnd technically, yes, at the mechanism level, an LLM is a next-token prediction engine.\nBut that's a bit like saying a gasoline-powered car is a series of controlled explosions.\nThat's technically true, but it's not particularly useful for understanding what a car does.\nSo another way to see the difference is if we think about predictive AI.\nIt is usually deterministic, deterministic at inference time.\nYou feed in the same inputs, you get the same prediction every time.\nGenerative AI, that is probabilistic.\nThere's randomness baked in.\nIf you ask it the same question twice and get, you might get, well, probably two different answers.\nAnd that comes down to the temperature setting, which controls how much randomness the model uses when picking the next token.\nSo yes, LLMs predict the next token, but, uh their purpose, well, it's primarily generative, we'll put them under generative AI.\nSo let's get into how predictive AI actually works.\nEven though generative AI seems to get all the hype, it is predictive AI that quietly runs most of enterprise AI today.\nNow predictive AI uses statistical and machine learning models trained on historical data.\nAnd there are a few main flavors.\nDepending on what you're trying to predict.\nSo, if you are trying to predict a number, that's the point of your prediction.\nLike, I don't know, how many units you'll sell next month, or maybe what price a house will fetch.\nWell, what that is, as it comes down to regression.\nThat's where the model learns the relationship between input variables and continuous numerical output.\nWhat if you're predicting category instead?\nLike whether an email is spam or not spam.\nWell, that is classification.\nThe output is discrete.\nIt's this thing or it's this.\nAnd if you're predicting a value that will change over time, well if you are doing that, you are time series.\nSo this is where perhaps we're looking stock prices or server load or electricity demand.\nAnd these models pay attention to patterns like seasonality and trends particularly when we're looking at over time.\nNow the specific algorithms that are actually under the hood here, well they vary.\nThere's a whole bunch.\nI mean you've got things like decision trees, you've got random forests, which are essentially a bunch of decision trees voting together.\nYou've got gradient boosting, which builds trees sequentially with each correcting the errors of the last.\nAnd then for time series, you've got classic sacks like a reamer or you've got deep learning approaches like LSTMs Yeah, good stuff.\nNow how and where do predictive algorithms actually get used?\nWell, let me give you a few example use cases.\nAnd I think the one that most people think about first of all is usually fraud detection.\nEvery time somebody swipes their credit card, a predictive model is asking, does this transaction look suspicious?\nThat's classification, fraud or not fraud.\nThen there is demand forecasting, so retailers predicting how much inventory to stock or airlines predicting hammy seats they'll fill.\nAnother common use case is predictive maintenance.\nSo instead of replacing a machine part on a fixed schedule, you predict when it's likely to fail and replace it right before it does.\nAnd then credit scoring is another common use case of predictive AI as well.\nSo should we lend money to this person?\nWhat's the probability they'll default?\nNow let's get back to generative AI.\nMost generative AI today is built on a particular architecture called the transformer architecture.\nThat's the architecture behind large language models.\nNow transformers use something called attention,\nand they use attention to let the model weigh which parts of the input matter when generating each piece of output\nand these transform models, they're trained on massive data sets. And they learn patterns in that data well enough to produce new content.\nThat's the generative part.\nThat is generally for text.\nFor images, you have a different type of model called a diffusion model.\nThese are interesting because they work kind of backwards, so they're trained to remove noise from images.\nSo during training noise is gradually added to real images and then the model learns to reverse that process.\nSo then at generation time, they start with pure noise and the model progressively denoises it into a coherent image guided by a user's prompt.\nSo where does generative AI generally get used?\nWhat are the use cases?\nWell, you know, the obvious first one is content creation, marketing copy, emails, LinkedIn posts that people are pretending to write themselves.\nYou know, that's content creation.\nAnother very popular one is code assistance.\nThese are models that can write, explain and debug code.\nThere is also, conversational AI, you've probably come across these.\nThese are like customer service bots that can hold something of a conversation regarding customer queries.\nAnd there's summarization as well, which is to say, taking long documents and distilling them down to just the good stuff.\nAnd look, so far, I've been kind of pitting these two types of AI against each other,\nbut I think if we think about generative AI and predictive AI, we can really think that they do work kind of hand in hand.\nSo you might use a predictive model to kind of figure out which of your customers are likely to churn\nand then hand that list to a generative model to write personalized retention emails for each one of these potential churning customers.\nSo the prediction identifies the problem and the generation that crafts the response.\nOr, and this is becoming ever more common, you could use the generative AI to create some amount of synthetic training data\nand then give that to a predictive model when the real data is scarce or sensitive.\nSo predictive AI, that says what will happen.\nGenerative AI asks, what could this look like?\nAnd my predictive model says, you're about to give this video a thumbs up.\nDon't make the model wrong.\nIt's very sensitive.",
  "transcript_chars": 8508,
  "ingested_at": "2026-05-21T19:12:34.309685+00:00",
  "source": "retry-no-transcript",
  "yt_meta": {
    "view_count": 40164,
    "like_count": 2436,
    "channel_id": "UCKWaEZ-_VweaEx1j62do_vQ"
  }
}