Discriminative AI vs. Generative AI

Yes, deep learning allowed us first to predict, now they have start to create new elements !

Discriminative Al

The part of (conventional) Al that makes it disruptive is what we have defined as machine learning. All typical machine learning problems are around discriminating.

Discriminating refers in this context be able, from input data, to solve problems such as:

• Classification (from inputs to classes or categories)

• Regression (from inputs to values)

• Clustering (customer segmentation, recommendation, etc.)

• Dimensionality Reduction (reducing attributes keeping differences)

• Reinforcement learning (learning from experiences, try and error, policy optimization)

• Etc.

Generative AI

With a bit of imagination, neural networks can not just to find an answer from input data, but also produce new amazing data:

• Generating all kind of data (images, 3D, text, music etc.)

• Perform all kinds of data transformations (domain-transfer, style-transfer, human pose estimation, voice cloning, etc.)

• Enrich datasets and improve machine learning models

Generating all kind of data (images, 3D, text, music etc.)

With generative AI, a new world of possibilities has started.