Generative models are one of the most exciting topics in deep learning nowadays. But with great power comes great responsibility. We can use the power of generative models for many good things, such as the following:
- Augmenting your dataset to make it more complete
- Training your model with unseen data to make it more stable
- Finding adversarial examples to re-train your model and make it more robust
- Creating new images of things that look like other things, such as images of art or vehicles
- Creating new sequences of sounds that sound like other sounds, such as people speaking or birds singing
- Generating new security codes for data encryption
We can go on as our imagination permits. What we must always remember is that these generative models, if not modeled properly, can lead to many problems, such as bias, causing trustworthiness issues on your models. It can be easy to use these models to generate a fake sequence of audio...
