Hidden materials on the back of creativity artificial intelligence
The original version from This story back Quanta magazine.
We have once promised robot cars and robot server. Instead, we see the appearance of artificial intelligence systems that can overcome us in the chess game, and analyze the huge texts for the composition of texts and songs. This was one of the wonders of the modern era: the physical work that facilitates human beings is very difficult for robots, while algorithms are increasingly able to imitate our minds.
Another surprise that attracted researchers in the long run is that this type of strange and strange creativity is.
Issue models, and the spine of photo makers such as Dall · E, Imagen and Pastable Diffusion to produce carbon versions of the trained images. However, in practice, they seem to wear elements in the pictures to create something new – not only colored spots, but coherent images with a semantic meaning. The paradox behind the version models, saying: “Giulio Perley, a scientist and physicist in SuPérieure école normale in Paris:” If they are completely, they should only remember. “But they do not do it – they are really able to produce new samples.”
To produce images, emissions models are used by practical emissions known as density. They convert a picture into digital noise (a group of pixels), then collect it again. It is similar to placing a plate through crushing over and over, so that all that remains is a soft dust candle, then connect the pieces together. The researchers were surprised for many years: if the models are rebuilding, how do you enter the image? It is like placing the chopped painting again in a completely new artwork.
Two physicists have now made an amazing claim: These are the technical deficiencies in the removal process that leads to the creativity of publishing models. In an article presented at the International Conference on ID 2025, this double developed a sporting model for trained publishing forms to show that what is called their creativity is in fact a specific process-a direct and unavoidable result of its architecture.
By running the black box for publishing models, the new research can have severe consequences for future research of artificial intelligence – and perhaps even for our understanding of human creativity. “The real power of the article is that it creates very accurate predictions of something very illegal,” said Luka Ambrogen, a computer world at Radoud University in the Netherlands.
under
Masson Kemboudi, a graduate student studying applied physics at Stanford University and the main author of the new article, has long been fascinated by formation: a process that has their own life systems.
One way to understand the growth of embryos in humans and other animals is named after what is known as the Torring style, a mathematician in the twenties in the twentieth century Alan Torring. Torring patterns explain how cell groups can organize themselves in separate organs and organs. More importantly, this coordination is local. There is no CEO for cell monitoring to ensure all matching with the final body program. In other words, the solitary cells do not have several body designs to create their work. They only behave in response to the signs of their neighbors and make reforms. The system is usually low, but for example, it becomes clumsy sometimes.
