Med-PaLM 2 is a variant of PaLM 2, a broad-language artificial intelligence model designed to answer medical questions and underpins Google’s Bard chatbot. It was announced at the Google I/O conference in May of this year.

what is known

The Wall Street Journal reported that an internal email they saw said they believed Google’s updated model could be particularly useful in countries with “more limited access to doctors.” Well, help poor people get medical care if they can’t afford to go to the hospital.

Med-PaLM 2 was trained on a hand-picked set of demos from medical professionals that Google believes will outperform generalized chatbots like Bard, Bing, and ChatGPT at medical conversations.

Short presentation of Med-PaLM 2 in English: video

• The article also published by Google in May and Med-PaLM 2 still suffers from some accuracy issues, which we’re used to seeing in other major language models. The developers found many inaccuracies and irrelevant information in the AI ​​responses.
• However, on almost all other measures, such as evidence of reasoning, consensus responses, or no evidence of misunderstanding, the Med-PaLM 2 showed “more or less the same results as real doctors”.

Med-PaLM 2 is still in its infancy, according to Google Senior Research Director Greg Corrado.