Undermind is a newly released AI-based system that searches for published research. It's like Google Scholar or PubMed, but it goes far deeper and is much more targeted when searching for a complex topic.
Undermind uses a new agent-like AI algorithm recently invented by MIT scientists to perform a search similar to how a human spends time looking through the literature with multiple steps of iteration. This makes it MUCH better at finding precisely relevant results.
In terms of performance, Undermind has been benchmarked to be 10-50x better at finding relevant content than traditional keyword search and other modern AI-based retrieval methods for the typical searches researchers want to do (see our whitepaper for detailed analysis).
An example comparing the same search on Google Scholar vs Undermind:
Google Scholar (click link to see this search)
Undermind (click to compare same search)
A quantitive comparison of Google Scholar vs Undermind
An analysis over a large number of typical user searches shows that Undermind vastly outperforms traditional search for finding relevant content (see our whitepaper for details).
In this figure we show for the median user’s search topic, Undermind finds far more highly relevant results, and makes it easy for the user to tell which results will be helpful to understand their search topic.
Our product is fairly simple, we first have a chat feature:
The other features are all part of the search reports that Undermind then creates for you: