Homeopaths, after detailed case taking, usually ‘“repertorize” the case using software such as RadarOpus, MacRepertory, Vithoulkas Compass, etc., and finally consult a Materia Medica to confirm the remedy selection. There are some highly experienced homeopaths who have the gift of identifying the correct remedy without even repertorizing the case, but this is rare. In general, arriving at the correct remedy involves a well-defined process and can be quite challenging.
The current practice of consulting electronic Repertories and Materia Medica, in my opinion, is becoming outdated. The process can be time consuming, and to add to the problem, there are different “schools of homeopathy”, which when followed, can lead to different remedies for the same case. This is not good for the patient since it can delay the cure. Homeopaths, like experts in other domains, must take advantage of new technologies to assist them. Wouldn’t it be nice if we can input the patient’s case record to the computer, which then suggests a suitable remedy? In this context, we need to seriously explore “ChatGPT”-like tools that take advantage of Large Language Models (LLM) for “Generative AI”.
RAG (Retrieval-Augmented Generation) is becoming increasingly popular these days since it aims to improve the quality of response from an LLM by augmenting its existing knowledge with additional facts. This also helps prevent “hallucination”.
As someone who is passionate about homeopathy, I have been exploring this technology for a while now and I see a lot of promise. In today’s article I would like to share my experience in interacting with an LLM using RAG.
I have with me a handful of documents about remedies such as Ars Alb, Lycopodium, Phosphorus, Nat Mur, etc., and I uploaded these docs to my chat environment. The following is a question I asked the system and its response.
You can see that the system attempts to give an answer based on the documents I gave it. Here is an important point to keep in mind: The quality of system’s response is very much dependent on my submitted data, which means the more details I share with it, the better it can get. Here are more interactions:
Interesting, isn’t it? I then asked the system to compare the two remedies Ars Alb and Kali Carb:
The system has responded quite logically. What you should note here is that none of the texts that I uploaded has this direct comparison between the two remedies. The credit goes to the system for picking up the relevant information from my documents, and more importantly, presenting it in a coherent form!
As the above brief interactions show, it is possible to use “Generative AI” to assist homeopaths in their case analysis. Of course, there is a lot of work to be done to make this succeed in the real world.
In my view, the ideal scenario is to build a “domain-specific” LLM for homeopathy. To give examples from other domains, we have BloombergGPT for Finance, Med-PaLM for Medical domain, and ChatLaw for Legal domain.
Building a domain specific LLM for homeopathy requires an enormous amount of data and substantial computational resources. It is not easy, but this needs to be taken seriously if we are interested in improving the quality of the system. Ultimately, this will benefit both homeopaths and patients.
Have a nice week and a Wonderful New Year!
Interesting it certainly it is to ‘upload information’ and see it transformed into ‘knowledge!