The Generative AI Revolution 

Those readers who have followed our blog posts in GRASPnetwork over the years will know that there have been several posts dedicated to the subject of artificial intelligence (AI) from various points of view, including the ethics of AI and the applications of AI to the arts. Why so much time dedicated to AI in our blog posts? Certainly, one reason is that some of us in GRASP have a long history of interacting with the AI community. My colleague Guenter Koch and I hosted a series of television programs in 1984, one episode of which was entirely dedicated to artificial intelligence. And so, we felt that this long history with AI and our continuing interest over the years made us uniquely qualified to write with authority and foresight on the state of the art and evolution of artificial intelligence and its associated technologies.

How wrong we were. How wrong we all were. Yes, we were well aware that around 2010, the scales had begun to tip in favor of the particular branch of AI called machine learning. Yes, we were well aware that this had happened in large measure because of the rise of the Internet, which had rendered massive amounts of data available for “training” machine learning programs. Yes, we were well aware of the advances in computing hardware that had made it possible to analyze all of this data. And Yes, we were aware of the impressive applications for vastly improved language translation and product recommendations – and even for winning television quiz show contests against humans.

But suddenly, on November 30, 2022, ChatGPT was released. And rest assured: nobody saw it coming. They say now that there is “AI before ChatGPT”, and “AI after ChatGPT”. A way to describe this watershed moment is to characterize “AI before ChatGPT” as analytical, and “AI after ChatGPT” as generative. That is, before ChatGPT, AI basically analyzed existing content and then, for example, translated it or made recommendations based on its analyses. ChatGPT introduced us to the generation of new content by AI. In particular, this was new written content – but it also included new image content, new musical content, and more.

There’s not much point in going into the details of how the generative AI technology works – the large language models, the transformers, the contexts, the feature vectors and tensors – there is plenty to be found online. But there is a point in pausing to think about the implications of this generative AI revolution for the issues that we in GRASP care about. As the followers of GRASP well know, our founding principle was all about coping with the increasing dematerialization of our society. Generative AI is yet another contribution to the speeding up of this process. Whether or not the fully dystopian vision of vast job losses to machines will be fulfilled like so many predict, it seems clear that generative AI will be injected more and more into our everyday lives, replacing more and more human interaction. GRASP’s motto of “making the invisible visible” will only become more difficult to realize.

As we said earlier, this revolution caught everybody off guard. We noted in an earlier post that the European Commission had prepared the “AI Act” to ensure the ethical use of AI. Its first version was released in 2017, and contained no mention at all of generative AI. Upon the release of ChatGPT in late 2022, there was a rush to include new considerations covering generative AI in the Act before it was finally published in March 2024. And so it is in many areas: everybody is rushing to understand the implications of generative AI for their work and lives, and rushing to cope. Nobody knows for sure where it will lead (if anybody tells you they do know, they are either naïve or cynical); but we in GRASP will be among those who are following it closely to determine its implications for the issues we care about.