Google’s AI medical bot, Med-PaLM 2 is being tested at various hospitals, including the Mayo Clinic Research Hospital. According to a report by The Wall Street Journal, the AI bot has been under testing since April.
The bot is reportedly being developed for usage in countries that have limited access to doctors. Let’s take a look at its background:
What is Google’s Med-PaLM 2
Med-PaLM 2 is designed to accurately answer complex medical questions, assist research to get more medical insights about patients, and much more. This AI model is a variant of PaLM 2 which was introduced at Google’s I/O keynote event in May.
Google revealed in a research paper that the AI tool still faced some accuracy issues. In this study, physicians pointed out many inaccuracies and irrelevant information given by Med-PaLM and Med-PalM 2 compared to doctors.
However, this medical bot performed well in other metrics like reasoning and consensus-supported answers, giving correct comprehension. However, similar to other AI technologies, Med-PaLM 2 is also prone to AI bias. In fact, it is a challenge faced by popular AI bots including Google’s Bard and ChatGPT.
.thumbnailWrapper
width:6.62rem !important;
.alsoReadTitleImage
min-width: 81px !important;
min-height: 81px !important;
.alsoReadMainTitleText
font-size: 14px !important;
line-height: 20px !important;
.alsoReadHeadText
font-size: 24px !important;
line-height: 20px !important;
The same report also stated that the customers testing the Med-PaLM 2 bot will be controlling the data, which will be encrypted, and more importantly, Google will not have access to it.
In a blog by the tech giant, Google unveiled that Med-PaLM 2 harnesses the power of other LLMs. Also, this AI model is among the first LLMs to perform and score 72.3% in the MedMCQA dataset that contains medical questions by Indian competitive exams of AIIMS and NEET.
It is important to note that this AI bot has been trained on a curated set of medical expert demonstrations- Google thinks this will make this chatbot better for conversations in the domain of healthcare than generalised chatbots such as Bard and ChatGPT.
Even so, according to Google’s senior research director Greg Corrado, Med-PaLM 2 is still in its early stages. He said he would not want to rely on it for his own family but believes it expands the chances of AI taking places where it can be beneficial in healthcare.
WHO on the use of AI in medicine
Earlier this year, the World Health Organisation (WHO) cautioned about using AI-generated large language model tools (LLMs) in healthcare. They highlighted certain concerns about AI bias, which can lead to them giving inaccurate or misleading information. Also, the WHO suggested these concerns be addressed before adopting AI in routine health care and medicine.
Edited by Kanishk Singh
