Judeumeh's Blog

What makes such Generative AI applications as: ChatGPT, Dall-E, Codex, or MidJourney and Stable Diffusion so compelling?  Is it the seemingly miraculous leap in capability, application and user experience? Or is it something to do with the timing, (at the tail-end of a great pandemic), and/or the tantalising promise of even greater things to come? And what has all this got to do with Intellectual Property anyway? Well, why don’t you jump on my multi-carriage train of thought to find out…

Note: This is the first in a series of posts intended to explore the capabilities, implications and potential ways to adapt and take advantage of opportunities offered by Generative AI. The first carriage on my metaphorical train looks at current-state AI and the amazing capabilities provided by transformers and other generative models.

This is truly a wow moment in time – when anyone can use AI in a meaningful way to create new works and content. The plethora of good quality generative text and image applications signal a sea change in the scale, speed and production of creative content. How else would you describe the glut of commentary and, dare I say, outright hype surrounding this topic? For example, some have gone as far as calling it the foundation of our future creativity, and The World Economic Forum considers Generative AI to be a game changer for which  society needs to be prepared. 

Breaking New Ground – Undoubtedly, the sensational release and uptake of OpenAI’s ChatGPT with its ground breaking capabilities and applications make previous incarnations of natural language AI applications appear somewhat limited. Generative Pre-Trained Transformer (GPT), is a transformer neural network, and deep learning model, trained on vast amounts of (usually Internet) data, which has consistently delivered better and more convincing results in successive generations. The current GPT-3.5 version, soon to be overtaken by an even more powerful GPT-4, makes Marvel Ultron analogies somewhat inevitable.  Various versions of OpenAI’s GPTs power such things as code generation (Codex), text-to-image generation (DALL-e) and even text-to-3D models (Print-e).

Something for Everyone – Outside of ChatGPT, AI text-to-image models and applications such as Stable Diffusion (stability.ai) and MidJourney (midjourney) can be applied towards producing or enhancing images, art, animation, videos, games and architecture with incredible results. We’re only just starting to see what these things can be made to do. In addition to generative code, image, and text, there are applications for generative music, speech, games and video. There are also directories available for finding the right AI application or model for your projects or interests, as well as adult oriented / NSFW applications. For better or worse, the image-laden porn industry is right at the forefront of the charge to adopt AI-generated imagery.

A Triumph of the User Experience – In my opinion, tools like ChatGPT have started to bridge the chasm of user experience which historically made sophisticated AI applications inaccessible to most users. When an easy-to-use technology application captures the imagination it can gain unstoppable momentum towards mass adoption, as exemplified by the level of engagement and excitement from anyone who has come across ChatGPT. The simplicity of use is critical to this phenomenon. Wolfram Alpha and other sophisticated AI systems exist, some of which may provide more verifiably accurate outputs, but lack that hook of simplified accessibility that ChatGPT brings to the table – for now.

So it is indeed a wow moment for AI mass adoption, and just as mass-produced automobiles did a century ago, it brings huge opportunities and challenges that will impact this and future generations.

The second carriage on my train of thoughts will look at the implications of these altogether amazing developments, and examine the ‘so what’ question which ought to be on everyone’s lips right now.