Choosing a quantum chemical method suitable for your simulations is not an easy task, because you need to balance accuracy andcomputational cost requirements. Unless you use B3LYP all the time, of course.

Generally, the more time you spend, the more accurate results you obtain. The emergence of AI models changed this landscape as, in principle, you can find AI models that are very fast and also accurate. Despite lots of potential and excitement about AI models, we have to admit that all of them have some problems. Hence, our research community is churning out one AI model after another at a neck-breaking speed with the hope to patch some of the problems.

Pragmatically, many of the AI and even quantum chemical models should be just version numbers of a general framework. Same with papers, they can be updated with results from new models. That would reduce the fatigue from the inflation of papers, and alleviate stress on reviewers who are now hard to find as everyone is busy reading or writing yet another AI paper. I am digressing as surely, in the near future, everyone will keep doing the same thing to keep our careers afloat.

To show that it can be done differently, we put out the fundamentally new paradigm in quantum chemistry, where we provide the platform with universal and updatable AI models – UAIQM for short. Instead of writing a separate paper* on each of the dozens new AI models we have recently developed, we put them together in a library.

Each of those models are already state of the art in many domains but the biggest strength of our approach is that we provide the framework for upgrading the models with more usage.

For your current application, the platform auto-selects the best AI model. The selection is based on how patient you are, i.e., if you are ready to wait a month for your simulation to finish, for example, when you go on your summer vacation – go ahead and choose the most robust solution.

If you are in a hurry and would like to finish your thousand MD trajectories overnight – we also have a solution for you! The best thing is that those thousand trajectories would still be more accurate than with B3LYP.

In any case, after you submit your computational job, the platform will complain if the chosen AI model is uncertain. If no warning is issued – you can have a good sleep and do not worry about the quality of the simulations.

The great thing about the UAIQM platform is that if your calculations were uncertain one month ago, the platform might provide you with a confident solution the next time you try it. For this, we collect** your uncertain calculations and improve our models for your difficult molecule.

**Please check the privacy policy what you should do if you do not want to share your toys with other kids.