Large language models (LLM) like ChatGPT have taken the world by storm, creating endless opportunities for better and more advanced tools. However, their reliability for medical and health-related information is questionable. Unreliable health information can be dangerous, leading to misdiagnosis and unnecessary treatments.
This is where Google’s Med-PaLM comes to the rescue. It is an LLM aligned to the medical domain, which goes beyond a simple search engine, offering purpose-built capabilities to assist medical professionals and enhance patient care.
At its annual health event, The CheckUp 2023, Google unveiled its upgraded Med-PaLM 2 with higher accuracy and reliability.
Let’s understand Med-PaLM and how this tool by Google can revolutionise the way we access medical information.
Understanding Med-PaLM
Med-PaLM, short for Medical Pathways Language Model, is Google’s revolutionary medical question-answering tool to provide accurate and reliable medical information. Developed using Google’s powerful pre-trained LLMs and expert demonstrations, Med-PaLM can transform the way we access medical information.
Med-PaLM 2 is an upgraded version of Med-PaLM with an impressive accuracy rate of 85.4% (19% more than Med-PaLM) on the US Medical License Exam (USMLE) questions. It also outperformed human ‘expert’ test takers, validating its proficiency.
Panels of physicians and users have validated Med-PaLM’s usefulness and reliability in generating long-form answers to consumer health questions. However, they found its accuracy and capabilities remain inferior to clinicians.
Currently, Med-PaLM is not available for public testing, but you can access the research paper to gain more knowledge.
Potential Benefits of Med-PaLM 2 in the healthcare sector
Enhanced accuracy and efficiency
Med-PaLM 2 surpasses earlier AI models by going beyond pattern-spotting. With its advanced understanding of symptoms, examination findings, and complex reasoning, it can deliver accurate and precise answers for informed decision-making.
Time savings and improved patient care
Med-PaLM 2’s quick and accurate responses to medical queries can save valuable time for clinicians, allowing them to focus more on patient care. Streamlining workflows enhances patient outcomes and overall healthcare delivery.
Access to reliable information
Pre-trained using powerful LLMs and expert demonstrations, Med-PaLM 2 can generate comprehensive long-form answers to consumer medical questions and provide trustworthy information.
Revolutionise healthcare sector
As AI continues to advance, Med-PaLM 2 represents a significant step forward in accessing reliable medical information and supporting healthcare decision-making. It empowers medical professionals, improves accuracy and efficiency and enhances patient care, ultimately revolutionising the healthcare sector.
Evaluation and Limitations of Med-PaLM 2
Med-PaLM 2 has shown impressive performance on multiple-choice medical question-answering benchmarks, and its answers have been favourably compared to those of physicians. However, its safe and effective deployment requires further work.
To responsibly deploy Med-PaLM 2, ethical considerations must be taken into account. Rigorous quality assessments for different clinical settings with appropriate safeguards to mitigate risks are necessary.
The potential harms associated with using LLMs for diagnosis or treatment necessitate careful evaluation. Additional research is needed to address biases and security vulnerabilities inherited from base models.
What does Google say about Med-PaLM 2?
Here are some perspectives from experts at Google:
According to Dr Alan Karthikesalingam, a research lead at Google Health, says Med-PaLM 2 has shown promising results in providing detailed answers, occasionally outperforming clinicians in controlled examples. Nevertheless, he acknowledges the need for further improvements and continuous learning.
Considering the sensitivity of medical information, Karthikesalingam emphasises responsible and controlled deployment, suggesting it may take time before the technology reaches the public.
Highlighting the historical impact of AI tools in medicine and the goal of AI in supporting caregivers, Karthikesalingam envisions AI enhancing healthcare by giving caregivers more time and enabling them to provide increased accessibility, availability and humanity in patient care.
Aashima Gupta and Amy Waldron, Global Director of Healthcare Strategy & Solutions, Google Cloud, expressed their dedication to realising the technology’s potential in healthcare through responsible exploration with trusted healthcare organisations.
Vivek Natarajan, a research scientist at Google Health AI, credited the success of Med-PaLM to advances in technology and the team’s medical expertise. He says that the combination of powerful LLMs and domain knowledge allowed the AI models to align with the nuances and safety requirements of the medical domain.
These perspectives highlight the ongoing evaluation, ethical considerations and potential benefits of Med-PaLM in revolutionising healthcare.
How does Google plan to take Med-PaLM 2 forward?
In the coming months, Med-PaLM 2 will undergo further development and improvements as Google continues to advance its research. The model will be evaluated across various dimensions—safety, bias, and helpfulness.
Google also aims to explore the integration of Med-PaLM 2 with other modalities beyond language, such as dermatology, retina, radiology (3D and 2D), pathology, health records, and genomics.
Leveraging a multimodal version of Med-PaLM 2, which incorporates images from chest X-rays, mammograms, and other areas, the technology has the potential to assist doctors in providing better patient care.
To ensure responsible and meaningful deployment, Med-PaLM 2 will be made available to a select group of Google Cloud customers for limited testing. This phase will enable the exploration of various use cases and the collection of valuable feedback while prioritising safety and ethical considerations.
Parting words
While Med-PaLM 2 shows promise in accurately answering medical queries and benefiting healthcare professionals, further research and collaboration with the medical community are necessary for widespread adoption.
It is crucial to approach implementation cautiously, considering the complexities of real-world settings. Only through rigorous testing, evaluation and responsible deployment can we fully understand the impact and limitations of Med-PaLM 2.
The Medical Futurist, in their article about Med-PaLM, say:
As these models get better and better, the risk of missing care due to capacity shortages in healthcare will soon outweigh the risk of the algorithms being wrong. We will be better off familiarising ourselves with communicating with such an LLM algorithm—purely because long waiting for medical answers due to the lack of healthcare personnel will pose a higher threat.
What do you think? Let us know in the comments.
