“AI researchers will soon be essential members of every chemical research team”

Research in chemical industry is often tiring and needs a lot of time. The Norwegian startup Iris.ai has developed an AI engine that uses natural language processing to review large amounts of scientific texts like research papers or patents. As a participant in 5-HT’s startup bootcamp X-Linker, Iris.ai was awarded second place among ten selected startups from the fields of digital chemistry and health. In this interview, CEO and founder Anita Schjøll Brede explains how Iris.ai can make chemical research more effective and efficient – and what the future of chemical industry could look like with AI assistants being an essential part of every research team.

Iris.ai Team

What are the problems about chemical research that Iris.ai sets out to solve?

The access to scientific knowledge has increased radically over the last 50 years, and it has become practically impossible to make sense of this load of information. Especially in chemistry there are a lot of scientific publications. The problem in chemical research is twofold: either you are not able to find what you are looking for because you are looking for a needle in a haystack, or you know that the research would take so much time. Regular keyword-based search engines are simply not sufficient; we need smarter tools to actually dig into the content of the papers and patents.

How can Artificial Intelligence help to make chemical research better?

We have spent the past five years building a core AI engine that reads and understands scientific texts based on natural language processing. Now, we have specialized this engine on chemistry where it can be applied for different use cases. Up to now we have developed three tools: discover, identify and extract. The discover tool is for getting an overview of the field: it is a new way of doing literature reviews and mapping out interdisciplinary research around a chemical research question. The identify tool is for when you know what you need. Then, the tool helps you find bits of specific information from literature. For example, it can help finding novel applications for an existing compound by scouting scientific literature. This makes it possible to discover new business options. Finally, the extract tool brings it all the way down to the granular level, extracting every key data points from a set of relevant documents, e.g. all experimental data in a set of patents. A job like this can perfectly well be done manually, but it is tiring and takes a lot of time. Within a few hours, our tool can complete a task that a human researcher would need two months for.

Iris.ai Software

How can AI actually make sense of scientific texts?

One good thing about chemistry is that scientific publications in this field are written in a very specific and concrete language. This makes it easier to analyze them with AI methods. First of all, we work with text similarity. By identifying the most meaningful words in a text as well as contextual synonyms, hyponyms or topic words, we make out the contextual fingerprint of a text and match it with other texts with similar fingerprints. Furthermore, we identify parent/child concepts and their correlation, and we also analyze the text for causality. It depends on the individual research question which of these aspects is most important.

Especially patent research often consumes considerable amounts of time and resources. Can your tool replace a comprehensive patent research?

Our tool will not replace every single task connected to patent research, but it will replace the tedious part of going through all the existing patents step by step. For example, one of our clients used to have researchers extract every piece of key data related to experiments from 120 patents to compare their own material performance against their competitors’ material performance. With our tool, they only need to drop the relevant patents into a folder, and the system will scan the patents, extract the relevant information and automatically populate a spreadsheet. This saves them from two full person months of work – every time.

Why did you decide to specialize your AI engine on chemistry?

We have been developing the core engine of Iris.ai for almost five years now – this has been a lot of work and effort. Our tool is already successfully applied in Academia, for example in university libraries in Finland and Norway. One year ago, we decided to specialize the technology and move into chemistry. This is a very interesting field of research, and we saw great potential for tech startups because many chemical companies are on their way to digitalization right now.

What are the next steps for Iris.ai?

At the moment, we are at the proof of concept level – we are delivering proof that our tool actually produces valuable information for our customers. Although we are already collaborating with a first set of clients, we are also looking for new clients. As our core engine can be applied for different use cases, we are excited to talk to the industry and to see if there are any additional useful features that we could integrate.

You took part in this year’s X-Linker program – congratulations on the second place in our startup competition! What was your experience with the program?

It was great because it was very focused and niched. We had the opportunity to talk to many companies and start good dialogues with them. It was also great to get the confirmation from both the companies and the judges that our idea is important. Our reward, the admission to Web Summit in Lisbon in November, will hopefully help us to further extend our visibility.

What is your vision for the future of Iris.ai?

We want to build what we call the AI researcher who can do research for us and with us. The AI researcher will not replace the human researcher, but it will become an essential tool for the researching team. It can not only find the right literature, but also draw conclusions from existing texts and even build new hypotheses based on research. The AI researcher will help humans to be more effective and efficient in their research and to make sense of all the information that is already out there.

Is the chemical industry ready for this major shift towards Artificial Intelligence?

In general, the chemical industry is lagging quite far behind in terms of digitalization and AI. Some chemical companies have already done a lot of work in AI, and when we talk to innovation and digital managers, we receive very positive feedback, but for many companies it is still a long way to go. All technology might be disappointing the first time you interact with it, and in all big companies you will have internal forces seeking to prevent this kind of change. But now, in times of the 4th Industrial Revolution, every chemical company needs to focus on digitalization. It is difficult, but it is vital to secure that you are part of the future. Now is the time to make these changes.

Be a part of the Digital Hub Mannheim/Ludwigshafen:

Get talking to innovative startup and established corporates in our network.