Artificial intelligence (AI) helps doctors in many ways. It acts like a smart assistant for diagnosing illnesses. AI models study huge amounts of medical data fast. This includes books, research, and patient records. Over two years, closed-source AI models led the way. These models solved tricky medical cases well. Open-source AI, where anyone can see and change the code, lagged behind.

A new study from Harvard Medical School, published in JAMA Health Forum, changes that. Researchers tested the open-source AI model Llama. They compared it to GPT-4, a top closed-source model. The test used 92 tough cases from a medical journal. Llama matched GPT-4’s skill. It got 70% of diagnoses right, while GPT-4 got 64%. Llama even picked the right answer first more often. This shows open-source AI now competes with the best.

Why open-source AI matters

Open-source AI has big advantages. Hospitals can run it on their own computers. This keeps patient data private. Closed-source AI sends data to outside servers. Doctors and tech experts can also adjust open-source models. They tweak it for specific needs, like local patient care. Closed-source tools stay rigid. However, closed-source AI offers easy support and works well with hospital systems. Open-source AI needs more setup work from users.

Both types of AI learn from vast medical info. They spot patterns, like cancer signs or heart issues. In tests, Llama proved its strength. It handled new cases after its training well. For 22 fresh cases, it scored 73% right. GPT-4 trailed slightly. Doctors see open-source AI as a game-changer. It gives them control over tech use. Each year, many patients suffer from wrong diagnoses. Better AI could cut errors and costs.

In a Harvard Medical School press release, the researchers say AI can boost diagnosis speed and accuracy. Hospitals prefer data staying in-house, and the researchers value open-source AI for local use in-house. Smart AI use could save lives and money. Doctors must guide it to fit their needs.

Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter.