INTELIGÊNCIA ARTIFICIAL GENERATIVA NAS CIÊNCIAS EXATAS, BIOLÓGICAS E HUMANAS: UMA ABORDAGEM DETALHADA DA ENGENHARIA DE PROMPT

• DOI: https://doi.org/10.22533/at.ed.1661125170314

• Palavras-chave: Inteligência Artificial Generativa, Instanciações da IAG, Engenharia de prompt

• Keywords: Generative Artificial Intelligence, instantiations of GenAI, Prompt Engineering

• Abstract: Generative Artificial Intelligence (GenAI) has emerged as a disruptive technology across the exact, biological, and human sciences, revolutionizing research processes, methodologies, and practical applications. Central to this transformation is Prompt Engineering, a technique that consists of crafting structured, detailed instructions to guide generative models such as GPT, DALL·E, and Midjourney. In the exact sciences, GenAI assists in solving complex mathematical problems, automating formal proofs, and performing computational optimization, with prompts ensuring precision and coherence in responses while accelerating simulations and physical modelling. In the biological sciences, GenAI applications stand out in genetics, personalized medicine, and bioinformatics. Well‑crafted prompts facilitate accurate diagnoses, the generation of clinical hypotheses, and the development of new pharmaceutical molecules, optimizing the use of large biological databases. In the human sciences (humanities), GenAI is particularly relevant for generating personalized educational content, producing detailed sociocultural analyses, and creating texts that are emotionally and culturally contextualized. It also holds significant potential for the analysis and generation of legal evidence, where structured prompts support the production of legal documents, the systematization of evidence, and the interpretation of legal texts with ethical rigor and argumentative coherence. Therefore, Prompt Engineering is a strategic and essential approach for the success of GenAI, requiring a balance among clarity, precision, and creativity, and ensuring the ethical and effective operation of generative models across diverse domains of knowledge.