BioGPT, a Microsoft AI-model trained for biomedical tasks

Microsoft, currently well-known for the release of ChatGPT in late 2022, has recently released another AI-tool, BioGPT. BioGPT is a new artificial intelligence (AI) language model specifically designed for the life science industry and optimised for answering biomedical questions. 

BioGPT is a generative language model, trained using biomedical text data, including scientific publications, clinical notes, and drug labels. BioGPT can use this data to answer questions, extract relevant data, and generate text relevant to biomedical literature. Compared to ChatGPT, which is trained on general text data, BioGPT has a deeper understanding of language used in biomedical tasks, and can be better applied to areas such as drug discovery and clinical decision support. According to Microsoft Research, BioGPT performs at a level of human experts in the tasks tested and outperforms other general and scientific language models. 

Recently, AI has shown the potential in areas such as clinical trial patient selection, predicting drug development outcomes, and developing digital biomarkers, however, there are still some limitations associated with the platform. BioGPT is trained specifically using biomedical literature and hence carries the same limitations as Chat GPT and AI in general. There are concerns that the models will produce inaccurate text, without any references, potentially providing misinformation. Additionally, the biomedical information on which the algorithm is based could carry biases, causing the AI program to further exacerbate these. 

BioGPT is still in the early phase of development and has some limitations, however, the platform has potential to unlock unprecedented levels of insights in the future.