The difference between generative and classifying AI
Artificial intelligence (AI) has recently become accessible to everyone with generative AI. Super handy, as with a written or spoken text, you can instruct an AI tool to generate, for example, a text or an image. On the other hand, we’ve been encountering classifying AI—also known as discriminative AI—for much longer in daily life. What exactly is that, and how does it differ from generative AI? In this article, we outline the difference between generative and classifying AI and discuss what you can do with each.
What does generative AI do?
Generative AI is a form of artificial intelligence that can generate new content. Think of text, images, music, and even videos. Well-known examples of generative AI models are GPT and DALL-E. The acronym GPT stands for Generative Pretrained Transformer, which is used in ChatGPT and other AI writers, including those in Microsoft Copilot. Generative AI compares patterns and then generates new ones. Applications that use generative AI are initially trained with human input: input from more or less creative minds, from Shakespeare and Multatuli to the Reddit forum. The choice of AI model and the quality of your prompt — your instructions to the AI — determine the quality of the generated output. Use generative AI to generate royalty-free images or jingles, or as a starting point for a blog.
A few benefits:
- Save time
- Get inspired
- Focus more on the creative process
Examples of generative AI applications
- Text generation: One of the best-known AI writers is ChatGPT. Great for blogs, but also for ads and newsletters.
- Image creation: Midjourney and DALL-E help with visual campaigns or design projects. Use them in combination with Photoshop or Canva for the finishing touch or to add the right text.
- Music and sounds: Jukedeck is an AI tool now integrated into TikTok to generate music and sounds for TikTok videos. An AI tool like Loudly focuses specifically on generating music for social media, ads, and videos.
What is classifying AI?
Classifying AI doesn’t create new content; it analyzes data and makes decisions based on that data. A few commonly used models for classifying AI are decision trees and neural networks. These systems are trained on a dataset where each input has a label, so the model learns to classify new inputs correctly. For example, a spam filter learns during training to identify and block spam. The model is typically updated periodically, based on user feedback.
Classifying AI models are trained to categorize data into different classes or categories. The software in which this model is implemented then makes decisions based on the labeled data. The indispensable ‘human in the loop’ oversees this decision process and adjusts it when necessary.
Examples of classifying AI applications
Classifying AI is used in:
- Security: An AI security system detects suspicious behavior, such as people driving slowly past construction sites multiple times. When the system detects such behavior, additional security can be deployed to prevent theft.
- Facial recognition: Programs like dating apps use AI for facial recognition, for example, in apps for women who love women. AI can distinguish quite accurately between men and women.
- Medical diagnosis: Classifying AI has been very successful in identifying irregularities in X-rays and other diagnostic procedures. AI has been shown to be faster and more accurate than the human eye.
When to choose which AI?
Choosing between generative and classifying AI depends on your specific goal. For new and creative content, you use generative AI. For insights into your data, you use classifying AI. Classifying AI identifies risks, but also new opportunities.
When ChatGPT and other AI applications became available to the general public, expectations were high. ChatGPT would manage the entire company, solve problems, and maximize profit margins. We now know that such superintelligence doesn’t exist yet in 2024.
Smaller, specialized AI applications are easier to deploy than a general AI tool. AI needs context to generate the right output from the billions of data available to the model. This is exactly what specialized AI software does: it gives the AI model the right context to function, for example, as a romantic songwriter, an appointment planner, or an inclusive job ad generator.
The AI tools of the future will combine different types of AI. In marketing, you need classifying AI to categorize users and their behavior and build profiles based on that. It would be ideal if the same tool also uses generative AI to immediately generate the right, new, and creative communication materials based on those profiles. AI tools like DataRobot, ATLAS.ti, Power BI, and Qlik are positioning themselves as apps that can do this.
AI-driven software like Textmetrics specializes in the labor market and diversity & inclusion (D&I). This gives the AI model a clear framework to generate content for attracting and retaining employees. Try out the job ad generator. With it, you can quickly write a job ad that attracts the right candidates!