Diffusion models – ethical issues
In this section, you will learn about the main ethical issues associated with using DMs for synthetic data generation.
Diffusion-based generative models are emerging and powerful technologies. Thus, their pros and cons need to be considered carefully. Their advantages are huge for businesses, industry, and research. However, they possess dangerous capabilities that can be leveraged to cause harm to individuals, businesses, societies, and so on.
Let’s list the main ethical issues usually associated with generative models and, especially, DMs:
- Copyright
- Bias
- Inappropriate content
- Responsibility issues
- Privacy issues
- Fraud and identity theft
Now, let’s delve into some of the main ethical issues behind using DMs in practice.
Copyright
DMs are usually trained on large-scale real datasets. For example, DALL E 2 was trained on more than 650 million text-image pairs. Thus, obtaining permission...
