Large Language Models (LLM) are usually trained on broad datasets that due to their lack of definitions for specialist or niche terminology lead to factual inaccuracies.
Recently LLMs have been trained using domain-specific data that although much smaller, complete tasks within those domains more quickly than general-purpose LLMs.
Large Language Law has discussed the development of targeted AI products trained on LLMs using proprietary legal databases and search systems. Similarly, Bloomberg found that the complex and unique terminology of the financial industry is best served by an LLM specifically trained with financial datasets.
Bloomberg GPT is an LLM trained on four decades of the company’s market data, financial research and analytics to assist with with risk assessments, judge financial sentiment, and potentially automate accounting and auditing tasks.
Leave a Reply