ChatMDR 2.0 February 2024
Client version 2.0.0
This is a big one. ChatMDR 2.0 introduces GPT 4 Turbo, the most capable AI large language model in the world today. This major update was actually on our roadmap for a little later, but early testing revealed that it was so much better than the already great output of ChatGPT 3.5 that was under the original ChatMDR. We decided to pull out all the stops and bring the new model to the market now. Right now.
What’s new
The February 2024 release has a number of new features:
GPT 4 Turbo
Although it stopped being cool somewhere in the early nineties to label something as Turbo, we were very impressed with the results of this AI model when we applied it to the MDR.
It’s hard to explain the difference between two AI models without getting very technical, but if you are so inclined, here’s an attempt. For ChatMDR, here are the main benefits:
- Better reasoning capability
ChatMDR 2.0 is smarter than its predecessor and better able to think its way out of regulatory problems - Larger context
The new model can take on much more information, which means it can combine more data points and work with bigger sections of the regulation, leading to higher quality and more detailed output - Creativity without hallucination
ChatMDR 2.0 responds much better to creative requests, such as ‘write a rationale’. ChatMDR 1.x was more locked down to prevent ‘hallucination’ (making things up)
Is there a catch? Maybe two small ones. First of all, other than what the Turbo name would suggest, the more advanced model takes more time to process. You’ll be familiar with this if you’re a paying ChatGPT subscriber and you’ve used the ChatGPT 4 model there. The results are worth the wait though.
The other thing is that it’s much, much more expensive to run the model. Think a factor of 10 more expensive. That is why the GPT4 Turbo model is available to ChatMDR Plus and ChatMDR Enterprise users. But – do not despair if you’re in the free tier! To celebrate the launch of ChatMDR 2.0, we are giving a month of free access to ChatMDR Plus away.
Instructions on how to sign up here.
Sources reflect last question
The Sources feature was introduced in ChatMDR 1.1 and shows where ChatMDR found its information.
Previously, the sources would reflect the entire conversation. This would make it hard if you went all over the regulation and covered big topics like clinical evaluation or conformity assessment.
Now, each answer will have its own Sources panel that only reflect the sources used for that answer. So if you decide to ask about the PRRC in the middle of a conversation about biocompatibility, it will produce Article 15.
Recap of Sources feature: Under each answer, a small link is displayed. When clicked, it opens a list of sources. These sources themselves are also clickable. ChatMDR will then tell you what this article, annex or document is.
Formatting
Many of our customers use ChatMDR to generate text to quickly copy and paste into their Word documents.
As of ChatMDR 2.0, numbered and bulleted lists are nicely formatted. Headings, boldface and italics are also supported.
Copying and pasting just became a lot easier!
Bug fixes and data updates
Once again, we had some unwanted guests on board that crawled around the electronics inside ChatMDR. They were asked in a friendly but firm manner to flutter about elsewhere.
(in case you didn’t know: read about the history of the term ‘software bug‘)
Also, as is our custom, the latest MDCG guidances were added and old updated ones removed.
Enjoy your new ChatMDR!
