Subscribe
Sign in
Cure AI — Your companion for evidence based research
Cure AI combines AI answers with seamless navigation of 26M+ PubMed articles. With advanced filters and evidence ranking by journal quality, citations, and relevance, it helps researchers and students find reliable, tailored insights effortlessly.
15
Replies
Best
Used it for my ongoing university research related to Parkinson's Disease - it is somewhat helpful, but I second the points raised by Konrad in the comments below. With gradual improvements, I'm sure it will be a useful resource for researchers. All the best :) @jared_watson
@jared_watson After quick testing, here some feedback
* It doesn't seem to understand questions precisely (asking "What are the most promising approaches to cure Lyme disease?", it started by listing methods for prevention; asking "Please describe the pathophysiology of Lyme disease in detail", it gave a summary of Lyme disease, but not much detail about the pathophysiology)
* It doesn't allow follow-up questions
* It doesn't prefer new articles (in Auto mode) when this would be essential (for my first example question given above, it listed sources from 1998-2016, while there has been much research in recent years and PubMed gives articles from 2020-2022 as first results for "lyme disease treatment")
* Only shows 4 articles per question, clearly that's not nearly enough
* Answer generation is quite slow
* Sources pane and source preview windows are quite nice, but highlights in the article abstracts don't seem to be related to the user's question. Sources pane should be shown by default.
* It doesn't indicate which articles are open access (a filter for that would also be good)
* Previous questions and responses are not restored after new log-in
* The color theme is quite unusual, there should at least be the option for something more conventional
@konrad_sx Hey Konrad, thanks for giving the product a try. Firstly, this is still an extremely early version, we are working with professionals in the space to optimize it for their workflow as well as improve our evidence selection algorithm.
We have not yet implemented context based conversation. This is definitely high up in the task list for updates to Cure but up till now we have been focused on optimizing our evidence selection algorithm.
For the preference to newer articles this is actually a difficult problem. For a lot of scenarios newer articles doesn't necessarily mean that the source is better than another. Our evidence selection algorithm currently weights similarity to query, the journal the articles was published in, the number of times it has been cited, and the publication type when evaluating evidence strength. You can specify the years in which you want your evidence to be selected within the advanced search parameter pop up as well.
Each question will be evaluated using 2-10 articles. Because this is still an early stage product sometimes some answers are better than others. If you are unhappy with your response often times just running the query again or changing up the query a tiny bit will improve the results.
Your feedback has been helpful I appreciate your help! I have added the open access indicator idea from your comment to our back log and hopefully some of your other suggestions like maintaining chat history, different color themes, and follow up questions will be implemented in the next updates!
@jared_watson Thanks for your detailed answer!
Some things do work well already, and it has some really nice features. You'll sure be able to fix the current issues. I understand that it's very difficult to build such a tool and needs time.
You're clearly right that a newer source isn't necessarily better than an older one, but other things being equal, I think it's very clear that this should be assumed. So I'd suggest to give weight to newness also. Optimally, maybe the weights should be determined by AI based on the query.
BTW, did you see the very similar product launched today? https://www.producthunt.com/post... See also my comment there.
@konrad_sx Will take a look. Appreciate the feedback and for using Cure. Hope you stay around for future updates
Cool. I think that for academic research can be pretty useful.
@busmark_w_nika Thanks Nika! We think so too, hopefully we can build something useful.
Congrats on launch! Definitely gonna try this out as we work with a few papers, exciting!
Cure AI