On Tuesday, Google launched a new family of artificial intelligence (AI) models called Med-Gemini for the medical industry. These AI models are not yet accessible to the public. Still, Google has released a pre-print edition of its research paper outlining its capabilities and methodology. The company says the AI models outperform the GPT-4 models in benchmark testing. One of this AI model’s distinguishing qualities is its long-context capabilities, which enable it to process and interpret health data and research publications.
Also read: Google Search Speaking Practice Feature Uses AI
What to Expect from Med-Gemini?
The research work is in the preprint stage and has been published on arXiv, an open-access internet repository for scholarly papers. In a post on X (previously known as Twitter), Jeff Dean, Chief Scientist at Google DeepMind and Google Research, stated, “I’m very excited about the possibilities of these models to help clinicians deliver better care, as well as to help patients better understand their medical conditions.” AI for healthcare will be one of AI’s most important application sectors.
Know about the Med-Gemini AI models
All of the new AI models are based on Gemini 1.0 and 1.5 LLM. There are four different models: Med-Gemini-S 1.0, Med-Gemini-M 1.0, Med-Gemini-L 1.0, and Med-Gemini-M 1.5. All models are multimodal, meaning they can output text, images, and videos. The models are also linked to online search, which the company claims has been enhanced by self-training to make the models “more factually accurate, reliable, and nuanced” when displaying answers for complex clinical reasoning tasks.
AI model has been fine-tuned to increase speed during long-context processing
Furthermore, according to the business, the AI model has been fine-tuned to increase speed during long-context processing. A higher quality long-context processing would allow the chatbot to offer more precise and pinpointed answers even when the inquiries are not perfectly posed or when processing a significant document of medical records. According to Google statistics, these AI models outperformed OpenAI’s GPT-4 models in the GeneTuring dataset on text-based reasoning tasks. Med-Gemini-L 1.0 also achieved 91.1% accuracy on MedQA (USMLE), beating its older model Med-PaLM 2 by 4.5 percent. Notably, the AI model is not accessible to the public or for beta testing. Before releasing the model to the public, the corporation will most likely make more improvements.
