Predictive vs Generative AI: How They Work and When to Use Each

Predictive vs Generative AI: How They Work and When to Use Each

by Martin Keen

Formal & Physical Sciences AI — News & Applications UYComputer scienceUYQArtificial intelligence (AI)
🎙 Martin Keen 👥 1.8M 📅 May 11, 2026 ⏱ 11 min 👁 55K 📄 tutorial 🧭 2026-08-06
Available in: English (current) Français

Keywords

predictive AIgenerative AILLMmachine learningdiffusion model

Summary

The video, presented by Martin Keen from IBM Technology, clarifies the distinction between predictive and generative AI. Predictive AI answers ‘what will happen?’ by analyzing historical data to forecast outcomes, such as fraud detection or demand forecasting. It typically uses structured data and outputs numbers or categories, with techniques like regression, classification, and time series analysis. Generative AI, on the other hand, answers ‘what could this look like?’ by creating new content (text, images, code) from unstructured data, using architectures like transformers for text and diffusion models for images. The video addresses the common misconception that LLMs are both predictive and generative, explaining that while they predict the next token, their purpose is generative. It also highlights how the two types can work together, such as using predictive models to identify at-risk customers and generative models to craft personalized retention emails. The presentation is clear, with practical examples, and emphasizes that predictive AI currently powers most enterprise AI applications.

159 words

Critical Evaluation

The video provides a solid, accessible introduction to the differences between predictive and generative AI, a topic often misunderstood. The presenter, Martin Keen, is an IBM employee, lending credibility, and the content aligns with industry knowledge. The explanation of predictive AI covers key techniques (regression, classification, time series) and algorithms (decision trees, random forests, gradient boosting, LSTMs), which is accurate and well-illustrated with use cases like fraud detection and predictive maintenance. For generative AI, the video correctly identifies the transformer architecture and attention mechanism, and explains diffusion models in a simplified but accurate manner. The discussion of LLMs as next-token predictors is nuanced, acknowledging the technical mechanism while clarifying the generative purpose, which is a valuable point for viewers. The video also emphasizes the complementary nature of the two AI types, offering a practical perspective. However, the content is introductory and lacks depth in mathematical or technical details, which might be a limitation for advanced viewers. The sources cited are limited to IBM’s own resources, which are relevant but not exhaustive. The video does not address potential limitations or ethical considerations of either AI type, which could be a gap. Overall, the video is informative, well-structured, and reliable for its intended audience, though it could benefit from more rigorous citations and a deeper exploration of the subject.

217 words

Title / Content Match

The title accurately reflects the content, which systematically compares predictive and generative AI, explaining their mechanisms and applications.

Quality & Reliability

8/10

The video is produced by IBM Technology, a reputable source, and presents accurate, well-structured explanations of predictive and generative AI. It correctly distinguishes the two paradigms, explains underlying techniques (regression, classification, time series, transformers, diffusion models), and provides practical use cases. The content is technically sound, though it lacks in-depth mathematical derivations and does not cite specific research papers, but it is reliable for an introductory audience.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Concurring Sources

Dissenting Sources

  • Some argue that LLMs are both predictive and generative — The video addresses this viewpoint, explaining that while LLMs predict the next token, their purpose is generative, which is a nuanced position.

Contribution & Novelties

The video provides a clear and concise comparison between predictive and generative AI, clarifying common misconceptions and offering practical guidance on when to use each. It emphasizes that predictive AI is deterministic and used for forecasting, while generative AI is probabilistic and used for content creation. The explanation of LLMs as next-token predictors is nuanced, and the discussion of how the two types can complement each other is valuable.

Pour aller plus loin :

  • Transformer architecture — Provides a detailed overview of the transformer model, foundational to generative AI.
  • Diffusion models — Explains the concept of diffusion models used for image generation.
  • Machine learning — General resource on machine learning, covering predictive techniques.

113 words

Radar Profile

The radar profile shows high scores in information quantity, quality, and technical level, with a slightly lower but still strong reliability score. This indicates a well-balanced, informative video that is technically sound and reliable, though not extremely deep.

Reliability 8/10

💬 Sur les 0 commentaires analysés, aucune tendance n'est disponible.