Från kursen: Generative AI Tools for Productivity and Research

Lås upp hela kursen idag

Gå med idag och få tillgång till fler än 23 100 kurser, ledda av branschexperter.

Evolution of generative AI

Evolution of generative AI

- Generative modeling techniques with machine learning algorithms is not a recent phenomenon. What we have now is because of iterations over years. The early generative models between the 1980s and 2000s, like the restricted Boltzmann machines, RBMs, and variational autoencoders, were an improvement over previous models that were only capable of generating text, as they could generate images and even audio data. And then in 2014, generative adversarial networks, GANs, were able to create higher fidelity images using two neural networks called "generator" and "discriminator." They excel at image sentences and style transfer. And then in 2017, transformers were introduced. Transformers particularly excelled at sequence to sequence tasks like language modeling, human-like text in dialogue generation that better the age of generative pretrained transformer, GPT. The now and future is multimodal modeling. This means training on different data types for more immersive and interactive…

Innehåll