The neural network model was named Med-PaLM 2. It can analyze medical data, answer doctors’ questions in detail, and give detailed explanations of the results produced. To obtain a license, the artificial intelligence passed a test that determined the accuracy of the information provided.

How does Med-PaLM 2 work?

The company tested the neural network on a series of medical tasks and questions. According to Google, Med-PaLM 2 answered 85% of the questions correctly – 18% more than the previous version. The developers claim that Med-PaLM 2 is significantly smarter than similar models from third-party developers.

However, neural networks still have a lot to learn. Developers continue to explore the capabilities of Med-PaLM 2:

• Participation in serious discussions that require a large amount of knowledge.
• Searching for information in complex and unstructured medical texts.
• To be able to summarize the information obtained from documents, scientific studies and information obtained during training.

Med-PaLM 2 was presented at the annual The Check Up event and is one of Google’s few advances in medicine. The company announced a collaboration to use AI technologies for Right to Care, a nonprofit. It monitors tuberculosis disease using neural networks in sub-Saharan Africa.