The recent advancement of deep learning and large language models has had a profound influence on many fields, including finance. Speakers Nino Antulov-Fantulin and Petter Kolm provided an overview of recent ML advancements like large language models, transformers, physics-informed neural networks, graph neural networks, and their impact on decision-making, data-driven analysis, and time series forecasting in finance.
The presentation paid special attention to large language models for finance. LLMs like GPT-3 are trained through a two-step process: pre-training and fine-tuning. In this pre-training, the model is exposed to a massive dataset of texts and is trained to the next word in a sentence or filling in gaps in sentences. In essence, the model learns to understand the grammar, context, and various language patterns. After pre-training, the model is fine-tuned on specific tasks or datasets to specialize its capabilities.
