Demystifying Generative AI Series: What is Generative AI?

How Does Generative AI Really Work?

In contrast to our ‘knowledge bases’ of the past, Generative AI algorithms are trained on huge datasets to produce new content rapidly. Content such as videos, images, text or music.
It’s versatility across domains and industry’s is remarkable – it leverages neural network architectures such as GANs and VAEs to do this.
GANs (generative adversarial networks) use a ‘generator’ (artist) – ‘discriminator’ (detective) model. VAEs (variational auto-encoders) is a neural network architecture, that helps generate new data similar to what it has been trained on. It uses a special technique with probability distributions to create this data. VAEs are often used in tasks like generating images or text and are useful in machine learning for making new content.
Using this iterative refinement process leads to continuous improvement of generated results.
Currently our corporate data is siloed. Data (accessible knowledge) such as LMS, sharepoint sites, excel, word, videos etc – are stored in silos. With AI – domain artifacts can be created across all data sets, thereby increasing employee productivity 5-10X.