Auto-translation used

NASA physicist Kyle Cabasares: The neural network completed the work in an hour that would have taken almost a year

NASA physicist Kyle Cabasares used artificial intelligence technology in scientific work. The results exceeded expectations. Cabasares shared with reporters that the OpenAI ChatGPT o1 neural network recreated the code of his doctoral thesis in 60 minutes. The physicist used 6 queries. A NASA researcher was surprised that the neural network was working so fast. For comparison, he added: it took him about 10 months to develop the code. Kabasares explained the agility of the neural network by the fact that it used its own synthetic data that it was asked to create, and not real astronomical data. The present work uses real astronomical data.

The physicist named the limitations he encountered during the study of the neural network. He says that for the program to work properly, additional input data and processing are needed, which must be performed using other software. The scientist came to the conclusion that the generated code can only be used as a basis, which requires improvement. This proves once again that artificial intelligence, despite its rapid development, still cannot replace the work of a real specialist.