The creators of project Aiur, an AI science assistant, envision a world where validated and reproducible scientific knowledge is available at our fingertips and is freely accessible to all to fuel core scientific breakthroughs. Jacobo Elosua explains.
What’s the main purpose of Project AIUR?
Our focus is on scientific research, specifically on chemistry. We’re trying to help different organisations, companies in the R&D sector and researchers to overcome the problem of identifying knowledge in an ever-growing ocean of information.
Chemistry corporates today spend, on average, two person months of highly qualified research personnel time extracting valuable scientific data from every hundred and twenty patents and papers pulled out of the tens of thousands deemed potentially relevant. In making this investment, those agents driving innovation forward still face very significant reproducibility and ‘unknown unknowns’ issues. These factors represent a massive inefficiency that current state-of-the-art artificial intelligence systems can help humans address.
We are building a knowledge validation engine that can validate research articles. Aiur gives a structured space of knowledge and tackles information overload. With Aiur you can quickly check hypotheses and identify research which is valid and relevant to your work.
The long-established procedure of scientific peer review means reviewers give their free time, and are not paid to validate scientific articles. We’re trying to help them by giving them a tool that allows them to do a proper peer review and make informed decisions.
How has your project evolved through NGI support?
With the financing and support from NGI, we are creating an incentivised scheme for authors and reviewers to increase the quality of research search results. And we want to help businesses gain more trust in the scientific community. Through the LEDGER program, we turned our attention to reproducibility issues, specifically in connection with machine-powered knowledge validation,
Through its validated knowledge-dependency trees, minimal viable project, the marketable version of Project Aiur aims at providing a critical first step in a long journey to optimize the combination of human and machine capabilities processing the vast and fast-growing volume of published knowledge currently locked in silos. A user will input a scientific document and the Aiur system, on the back of having processed millions of filed patents and published papers, will output a dependency tree for human researchers to validate. This inter-document dependency tree will highlight key building blocks on which knowledge is advanced, helping researchers identify corroborating and contradictory evidence, as well as missed constraints. Post validation, the individual knowledge tree will be merged into an ever-more-intelligent chemistry knowledge graph.
What’s your background?
As European co-founders who first met in the US at Singularity University, we dug deep to figure out what was the impact-driven mission we would devote the next ten years of our lives to. As the conversation evolved, it became clear that the four of us had faced deep personal frustration from the inability to capitalise on scientific knowledge.
We created Iris.ai and beyond the fast-growing movement to – rightfully – make Open Access research the new norm, the premium versions can be connected to any scientific information repository a user has a license for. In 2015 Iris.ai started a long journey to develop the world’s leading Artificial Intelligence Science Assistant. We’ve been featured in Forbes magazine and were selected as a Top 10 contestant in the global AI X Prize.
We are self-described ‘geeks’ and as we developed Iris to apply it to chemistry research, we decided to name the project Aiur, after a planet in the StarCraft II sci-fi video game.
What does the future hold for Aiur?
Right now, we are excited about the very promising dialogues we have kickstarted with some of Europe’s leading research institutes. Over the coming months, we hope to incorporate their feedback into our product development roadmap, to be able to furnish them with a tool to improve the quality of their early stage research processes. We will also resort to distributed ledger technology to formalise multi-party knowledge management protocols, as well as, eventually, to address current built-in biases and incentive misalignment issues.
Project Aiur represents a critical step in our journey to develop the world’s leading Artificial Intelligence Science Assistant. So if you read this and you think we should talk about any potential collaboration around knowledge validation, please do drop us a line at founders@iris.ai. We would love to hear from you!