The Future Is Going Extinct
What happens when we give up our ability to dream?
An unstoppable movement has begun. The arrival of artificial intelligence that is available to everyone, along with the opening of code so that anyone can experiment on their own launches us into a process that has the power to change our relationship with the future. The opening of GPT-3 to the public use in late 2021 was a lead-up to the breakthroughs in image creation we saw since August 2022, and this revolution soon spread to the audiovisual world, code, our everyday language, and potentially all other areas of our lives. We are experiencing the culmination of a dynamic we saw coming: the ability of more and more people to access a powerful tool whose impact is difficult to imagine. AIs such as those already in continual evolution, with learning potential, speed, the ability to transmit emotions, and the tools to create new elements, will not only provide us with beautiful and unexpected products but will also revolutionize the depth of what we know. Intelligences like these could wipe out our future.
Outsourced imagination
We ask image-generating artificial intelligences to dream, to imagine with us. This is how we code our requests in most systems: “dream about…” (hence the name of the stability.ai project: DreamStudio). The results are images with impressive artistic quality that we cannot distinguish from human creations. These images, among others, renew debates about what art is and is not, and who are artists or not. AIs learn from us and with us, based on all the human knowledge we store in the Digital Environment. While most people argue about instrumental consequences (how to assess learners, how to plan education, how to distinguish true from false), my question is: are we delegating our ability to imagine?
Dream about the end of the human race. Dream about the date of environmental collapse. Dream up the answers to questions we don’t dare ask each other, dream of the forbidden. What will happen when AIs dream of unknown events, when they can tell us about something with such a degree of certainty that they “believe in” the reality they’re about?
Artificial intelligences have made remarkable progress in recent years. For example, the company DeepMind, known worldwide for creating AIs with the ability to beat the world’s best Go players, now creates AIs that allow us to perform scientific feats never before imagined. “Dream, AlphaFold, about the structure of proteins. Predict a result that will help us in our scientific pursuits.” The challenge of knowing the structure of a protein is crucial to many aspects of medicine and life sciences. For decades, work has been done on the experimental capacity to determine that structure from the amino acids that make it up, to predict its shape based on its basic components. The problem is that experimental work takes time. AlphaFold has shown that it can predict the structure of 20,000 human proteins and 20 other organisms with an accuracy close to our best experimental methods, but in seconds. This is surprising, but not unexpected. What is unexpected is that it happened in 2022, when it’s the sort of thing we have always imagined as our future. As 2009 Nobel Laureate in Chemistry, Venki Ramakrishnan, stated, “It has happened much earlier than many experts had predicted.” There’s a fine line between dreaming and predicting.
Is the future still unknown?
NVIDIA’ Earth 2 project is advancing and getting more resources to monitor and predict our planet’s climate. Soon, AIs will inform us about what our world is like, what it could be like, what it will be like. So I come back to my questions: what will happen when we can foresee something with the level of accuracy achieved by AlphaFold? We would be making irrefutable assertions about the future… But then, the future would be the present…
If someone decided to ask an AI about the end of humanity on planet Earth and got a calculated answer, would we be able to continue living with that global expiration date in mind? What would happen to the way we care for the environment? What would happen regarding the management of resources?
The evidence indicates that we would be catapulted into the end of the world at almost the same moment the prediction was made. The historical debate over free will versus destiny becomes relevant again, though in an updated context. Do AIs have the capacity to condition us, to reduce our field of action?
Giving up the ability to imagine can be dangerous, but it’s also deeply distressing. And let’s not forget that AIs make their predictions without the evolutionary burden we have as humans. We may not be ready for those answers. Those outputs may be so far outside our capacity for imagination that they annihilate it.
Stability.ai claims to create open digital intelligence tools that allow us to reach our true potential. That all sounds very nice and hopeful, but it encapsulates an inescapable question: what is that true potential of ours? How can we measure it? How can we recognize it? What does it mean to be human beings on the verge of the singularity?
For the time being, we run the risk of surrendering our future, of wiping out that unpredictable development, the not-knowing that keeps us imagining.
