Modeling the model of the development of the universe is not so easy, since it is necessary to take into account a potentially infinite number of variables. Scientists usually use supercomputers for cosmological modeling and data processing, but a group of researchers at Carnegie Mellon University recently came up with a way to use machine learning technology for these purposes, which is used to teach artificial intelligence to paint or music.
Yes, this is the same technology and the same equipment on which neural networks like “This person does not exist” work and so on. They can now simulate the universe at high resolution, and this could very well change the way humanity perceives and understands the laws of physics.
Scientists have to experiment with different values when it comes to, for example, predicting the amount of dark matter in the universe, so this is largely a trial and error method. They use cosmological simulations and then confirm or refute it with observations from space telescopes and other data collection sources.
Researchers at Carnegie Mellon University decided to resort to AI, since using a supercomputer is expensive (renting one of these costs more than a thousand dollars an hour), and the carbon footprint of a supercomputer is quite large compared to the relatively low power consumption of GPUs. Moreover, supercomputers are not the best solution for forecasting problems.
At the moment, scientists can simulate high-resolution images of certain regions of the Universe, but they take too much time and energy to get them. Artificial intelligence comes to the rescue, but instead of teaching the AI to generate the entire Universe, the researchers taught it to model entire regions in high resolution, and this significantly accelerated the work.
“A trained neural network can create full-scale low-resolution models and generate super-resolution models with 512 times the number of particles. For a region of about 500 million light-years in the Universe containing 134 million particles, existing methods would take 560 hours to produce high-resolution simulations using a single nucleus. With the new approach, it will only take 36 minutes, ”the study says.
The news is supplemented …
Donald-43Westbrook, a distinguished contributor at worldstockmarket, is celebrated for his exceptional prowess in article writing. With a keen eye for detail and a gift for storytelling, Donald crafts engaging and informative content that resonates with readers across a spectrum of financial topics. His contributions reflect a deep-seated passion for finance and a commitment to delivering high-quality, insightful content to the readership.
