Google’s medical AI bot, Med-PaLM 2, is undergoing testing at hospitals, including Mayo Clinic Research Hospital. The AI bot has been under testing since April, according to a report by The Wall Street Journal.
Developers are working on Med-PaLM 2 for countries with limited access to doctors. The testing process of the AI bot is being conducted at various hospitals. The Wall Street Journal reported on the development and testing of Med-PaLM 2.
Med-PaLM 2 accurately answers complex medical questions and assists in patient research. PaLM 2, introduced at Google’s I/O keynote event in May, serves as the base for Med-PaLM 2.
Med-PaLM 2 showed good performance in metrics such as reasoning and consensus-supported answers. AI bias is a challenge faced not only by Med-PaLM 2 but also by other AI bots like Google’s Bard and ChatGPT.
According to the report, the customers testing Med-PaLM 2 will have control over encrypted data. Google will not have access to the data used in the testing of Med-PaLM 2. Based on the report, the performance of the medical AI bot in comprehension was deemed correct.
AI bias is a known issue that Med-PaLM 2 and similar AI technologies are prone to. However, the report highlighted that customer data for Med-PaLM 2 would be controlled and encrypted. Additionally, Google’s access to the data used in testing Med-PaLM 2 was explicitly stated as restricted.
According to Google’s blog, Med-PaLM 2 utilizes other language models (LLMs) capabilities. Remarkably, Med-PaLM 2 scored 72.3% in the MedMCQA dataset, which includes medical questions from Indian competitive exams.
The AI bot was specifically trained on a curated collection of medical expert demonstrations, making it more suitable for healthcare conversations. In contrast, general chatbots like Bard and ChatGPT are not as specialized for healthcare discussions.
Greg Corrado, Google’s senior research director, stated that Med-PaLM 2 is still in its early stages. Despite expressing caution and stating that he wouldn’t rely on it for his own family, Corrado believes that Med-PaLM 2 expands AI’s potential benefits in healthcare. The blog post emphasized the fact that Med-PaLM 2 taps into the potential of LLMs.
It highlighted the impressive performance of Med-PaLM 2 in the specialized MedMCQA dataset. Overall, Google sees Med-PaLM 2 as a step forward in enabling AI’s positive impact in healthcare.
Google Still Working on AI bot’s Accuracy
In a research paper, physicians noted inaccuracies and the presence of irrelevant information in the responses provided by Med-PaLM and Med-PaLM 2. Google acknowledged that the AI tool still faces accuracy issues based on physician feedback. Med-PaLM 2 aims to improve medical insights, but it encountered concerns about accuracy in the study.
WHO raised Concern
Concerning the use of AI-generated large language model tools (LLMs) in healthcare, the World Health Organization (WHO) issued a warning. The WHO expressed concerns regarding AI bias potentially causing inaccurate or misleading information in healthcare. Before incorporating AI into routine healthcare and medicine, it is recommended to address these concerns.
On the other hand, the WHO emphasized the need for precautionary measures regarding AI bias in healthcare applications. Their caution aims to ensure the accuracy and reliability of AI tools in the medical field.
More About the Usability
Med-PaLM 2, a specialized version of PaLM 2, demonstrated state-of-the-art performance in answering medical exam questions. Thorough human evaluation is being conducted to explore Med-PaLM 2’s potential in assisting healthcare organizations.
The AI model aims to draft responses, summarize documents, and offer insights to healthcare providers.
However, developing accurate AI systems for medical questions has been a long-standing challenge. Nevertheless, Med-PaLM 2 shows promise in addressing this challenge and supporting healthcare organizations.
Furthermore, Google Health is actively developing and testing AI models to address global physician shortages and limited access to medical tools. They aim to improve accessibility and provide patients with timely, accurate diagnoses and care.
Through enhanced technology, the objective is to help patients receive necessary medical attention. Specifically, the focus is on alleviating healthcare disparities and improving healthcare outcomes globally.
Ultimately, the aim is to leverage AI to bridge gaps and enhance healthcare accessibility for underserved regions, benefiting needy individuals.
Sequencing genomes helps identify DNA variants indicating genetic disorders like elevated breast cancer risk. DeepVariant, an open-source variant caller, utilizes a deep neural network to detect genetic variants. Next-generation DNA sequencing data is used to call genetic variants using DeepVariant.
Exceptional Research Ability
Google Health offers secure technology to healthcare professionals for conducting research and improving health understanding. Researchers interested in collaborating with Google Health for health research can provide their details. Google Health will notify interested researchers when research partnerships become available. The focus is on facilitating research collaborations to enhance healthcare outcomes and knowledge.
Ability to Early Intervention
The research in Nature showcased AI’s ability to predict acute kidney injuries (AKI) earlier. AKI, challenging to detect, affects many hospitalized patients in the US and UK. The AI model accurately predicts AKI up to 48 hours before the current diagnosis. The research highlights the potential for early intervention and improved patient outcomes in AKI cases.