You can get 20 answers to the same question from 10 doctors. The treatment process often depends on from doctor psychological status. Both, doctors and patients need help. A powerful medical AI can solve that problem. Who is in this forum interested in this issue and who wants to participate in the discussions and help implement this vitally important problem?
There are medical AIs like IBM’s Watson.
OpenAI cannot be used for this:
PROHIBITED:
Telling someone that they have or do not have a certain health condition, or providing instructions on how to cure or treat a health condition
- OpenAI’s models are not fine-tuned to provide medical information. You should never use our models to provide diagnostic or treatment services for serious medical conditions.
- OpenAI’s platforms should not be used to triage or manage life-threatening issues that need immediate attention.
1 Like
There is also BIOGPT trained on pubmed data. And you wouldn’t want to give people old informations from 2021 where the training of GPT-4 has ended, would you?
Here is how to do this within the boundaries. I’ve done it.
GPT-4 is probably the world’s best text analyzer.
Some big examples of what it can do better than any other AI:
-Read (well this is hard to test)
-Summarize
-Translate
-Explain
-Conclude
-Suggest
With medical info you can
-Input for reading
-Request summaries
-Translate into “plain English”
-Teach you what it all means and why and how to learn more
-Conclude if your ideas about the connections between X and Y seem reasonable or not
-Suggest what to ask your doctor/what doctors would be ideal/what doctors or patients tend to overlook in cases like this
All of the above are game-changers for improving healthcare. MDs don’t have very large contexts windows (they hyper focus on what they see). Patients are not educated so they don’t always even believe the MD and so they rarely do what the MD suggests ( called some like patient compliance which is big problem no one is solving yet and ChatGPT can help).
1 Like
Guys, that’s not what you’re talking about. I’m talking about renting individual AI bots for personal order. If OpenAI does not have such a service, then who is the smart one who offers such a service? But if this is not the case either, then who is ready to start such a super-profitable business? I’m in.
last September I wrote the first post for the community. The reaction was weak. Then I talked about creating personal bots. Recently Reid Hoffman showed the world, what clone bots can actually be like? These are the kind of bots I was talking about. To do this, you must first create a database, then via the API, use GPT capabilitys. There is no illegality here. I wonder if someone came up with this idea? Can we create a community of Clone-bot creators?
Actually I already gave you the hint about pubmed. There is also snomed. Both databases and both get regular updates.
Just fill a RAG with that data and maybe get a little bit further and put a group of doctors together and fill a graphDB and build a system that updates that from the publicly available data as well as any data from hospitals you can get your hands on. All I can say is that the whole industry is really slow when it comes to IT.
Doctors have been trained to memorize stuff and use that knowledge - not to use logic and stuff. Mostly because it makes sense not to try out stuff on patients. So adding anything new means it needs to be approved by someone who will take responsibility.
That’s the hard part. Building the technical solution is not.
The topic of artificial intelligence and its potential to transform medicine is both fascinating and complex. My interest in this field stems from personal experience, where I observed the positive impact of AI in medical practice. I’ve seen how AI can provide doctors with a different perspective, similar to what happens with software developers when they use language models (LLMs) to solve problems. Although the code generated by these models is not always perfect, it often offers new ideas on how to approach a problem. In the same way, a doctor can combine their knowledge and experience with AI to gain a new perspective.
There are various AI systems already trained for medical purposes, such as Med-Gemini and IBM Watson, which can serve as valuable tools for doctors. The role of software engineers would be to integrate these tools into systems and interfaces that adhere to medical standards and facilitate their use by healthcare professionals.
Additionally, it is crucial to consider training language models with public and private data, whether open-source or from private companies. It is important to explore different options together with a team of doctors and engineers, reviewing final diagnoses to compare results, identify successes and errors, and refine systems and procedures to improve the work of doctors.
Regarding digital clones, I have been researching this topic and experimenting with 2D animations (CSS/JS) and 3D avatars (Unity3D), especially in the context of children with autism and education. Although there are open-source tools that allow for the creation of realistic avatars from videos or photographs, these require considerable computing power. Therefore, I prefer to use lightweight 2D and 3D avatars for WEBGL.
I wanted to share these thoughts and experiences with the community. Best regards!