AI assist MO’s for best treatment plans

[fmenninger]

As we have a conundrum of too many treatment options but which one to choose for best outcome for me? It’s obvious that pca is very complex and what works for one May or may not work for another. So, can MO’s carefully use chatgpt and AI to scour data to come up with best treatment plans or is this too far fetched? How accurate would this be when comparing AI vs a top notch MO? Not to suggest that AI will replace doctors, of course, only to help in collecting and analyzing copious amounts of data. Thoughts, ideas and suggestions……

[tango65]

I checked those systems and they can present a logic and sometimes correct assessment of a medical condition but they fail completely when one ask them for the links to the articles they used to come up with the assessment.

All the links were wrong, even the title of the articles were wrong or they don't appear in any search in PUBMED or google. So, I won't trust these AI systems yet.

[Nusch in reply to tango65]

ChatGPT has no direct access to internet, therefore most of the links and sources provided are wrong. Pls note as well, that ChatGPT3 can only access information up to 2021. But you can get access to ChatGPT4, which is up to date and is an improved version. Using ChatGPT3 is free, for ChatGPT4 you have to pay a small fee.

If you are in the US, you can already access BARD, this is the equivalent AI tool from Google. As far as I know it’s free. It has some nice features, which ChatGPT doesn‘t provide at this time: you can send a PDF file, e.g. a complete study, to BARD and ask BARD for analyzing, summarizing and explaining the content. Works great!

In the next months and years we’ll see many many improvements, the progress will be exponential and it will help us all, the become a better pilot of our life and disease.

One example: they tested AI tools against a team of doctors when searching for tumors on a scan. The AI tool was much fast and more accurate While team of doctors found only 67% of tumors, AI tool found over 85%. They also let an AI tool do the tests for medical exams and the AI tool came up with tremendous results.

So the knowledge of AI will outperform the knowledge of one expert quickly. It will also help to invent new treatments and create individual therapies (which is especially important for immunotherapies).

I trust that we’ll see a „hockey stick“ development within the next years. So let’s do everything possible to survive at least five more years.

[tango65 in reply to Nusch]

The interesting thing is that when asked (bard and chatgpt), they provided links and when I told them all the links were wrong they apologized and mentioned they do not have direct access to internet. I believe in AI and I think it is already helpful in medical sciences

[Nusch in reply to tango65]

Maybe this helps: zdnet.com/article/how-to-us...

[Purple-Bike in reply to tango65]

I found the same when using chatgpt over openai, but after starting chatgpt4 on Bing I am so far - not done it a lot yet - getting correct links. It really seems to be superior.

[tango65 in reply to Purple-Bike]

Thanks . There is a big difference, and I believe is useful to finding relevant medical info if one knows what to ask for. Citations are correct.

[Purple-Bike in reply to tango65]

Indeed, a well thought out prompt, maybe a few follow-up questions, a Google Search in between and reading the most relevant linked study to arrive at some understanding of a complex medical issue to the best of one´s ablity - the technology is empowering.

[Purple-Bike in reply to Nusch]

I would gladly pay a fee for ChatGPT4, but I have it for free. Perhaps additional features come at a cost.

[Tall_Allen]

Not ChatGPT, but an AI approach has been tested for assigning risk groups at diagnosis and seems to work well.

artera.ai/

[SierraSix in reply to Tall_Allen]

a LLM specifically designed for PCa. I'm assuming it's already being worked on. Having a large language model(s)trained specific on PCa (and other cancers, disease...etc) should provide break thoughs in many areas of care

[SierraSix in reply to SierraSix]

btw I fed my MRis, pathology reports and psma scans as input into my gpt prompts and was able to summarize, compare findings. Etc. Would I base a treatment on it. Not yet. But AI will get very good at things like radiology. Etc imho

[fmenninger]

very interesting topic as this is really the tip of the iceberg as I am sure AI will get better but will have its pls e and won’t replace MO’s anytime soon. Can we train chatgpt to be an expert in apca?

[j-o-h-n]

Artificial Intelligence is the daily bullshit coming out of Washington D. C.

Good Luck, Good Health and Good Humor.

j-o-h-n Wednesday 05/17/2023 10:46 PM DST

[fmenninger in reply to j-o-h-n]

ha ha…so true. Here is a good read using AI to treat pca: blog.google/technology/heal...