While OpenAI’s ChatGPT has become almost synonymous for generative AI, there are many exciting projects and developments that can complement what Microsoft and OpenAI offer together quite nicely and that deserve some attention.
In this session, we aim to look beyond working with text, and beyond using cloud services. We will cover alternatives to ChatGPT, including Open Source models, and why you should care.
This includes some large cloud-hosted models, relevant platforms, and integrations of AI models in new types of tools. We will talk about how so-called “base models” are different from “fine tuned” models, and how a model becomes a ChatGPT-style model. We will look at relevant projects and training datasets and underlying approaches.
We take a closer look at what it takes to actually run Large Language Models on your own hardware, enabling scenarios where data cannot be sent to cloud services, or where contractual limitations may prevent specific use cases. Attendees will receive information how to get started.
We then move on to other so-called ”modalities”, such as generating images. Taking a closer look at Stable Diffusion and the rich ecosystem around it, we will cover briefly what Diffusion models are, and explore capabilities that go beyond what users can easily consume in the cloud. This goes from editing existing images, putting your own likeness into image generation, and utilizing image generation to explore some interesting underlying concepts, such as the “Latent Space”.
Next we aim to combine some AI capabilities such as so-called “AutoGPT” systems or “Agents” to do interesting things.
Finally we will talk about some of the latest innovations that we have found.
We will not forget Responsible AI implications when it comes to the discussed technologies.
You will learn:
- About generative AI complementing ChatGPT
- How to get started with running large models locally
- Hear about latest trends in generative AI