In this seminar, I will present my personal exploration of the application of artificial intelligence in the study of galaxy evolution. Since the early 2010s, the fields of machine learning and deep learning have undergone unprecedented development, marked by enhancements in algorithms and hardware advancements. This progress has significantly influenced various scientific domains. Concurrently, the increasing volume and complexity of data produced by modern and upcoming astronomical surveys such as J-PAS, DESI, LSST, and SKA necessitate the adoption of new analytical tools capable of processing and interpreting this data faster and effectively.
Initially, the first applications of AI in astronomy made used of multilayer perceptrons to address challenges such as estimating the photometric redshift of galaxies. Today, most AI applications in astronomy primarily employ supervised learning with sophisticated neural networks like CNNs, RNNs, and LSTMs. These are focused on estimating physical properties or performing classifications of astronomical objects. However, a shift towards a new paradigm in computer science emphasizes the use of generative AI models trained via unsupervised learning methods. Prominent examples include tools like ChatGPT, Stable Diffusion, Sora, Udio, and others. This transition is now making inroads into the field of astronomy, offering promising avenues for the development of what is being called foundational models. These models have the potential to integrate data across multiple surveys and wavelengths, creating a more robust framework for interpreting theory and observations.