Iris.ai - ten years later
It is almost unfathomable that we’ve been working on Iris.ai for a full decade. Reflecting on this journey will take way more than a blog post, but as the journey continues and we have work to do, the memoir will have to wait. The path has been full of twists and turns, and I sometimes think of our Iris.ai as a cat - she has nine lives, of which we’re now in something like the fifth. Yet, there are some things that are permanent, and today there are a lot of our efforts converging into a very, very exciting market position.
Here are my main takeaways from the road thus far - and for the next decade:
We were so young, a tad bit naive, and early 10 years ago
When I stood on the TEDx stage in a dutch castle in early 2016, announcing to the world that we were building what we then called “An AI Researcher”, I think most people thought I was a little crazy. We were Norway’s first AI Startup, we were fully remote with co-founders living in four different countries, we had gotten our start at the visionary and sometimes mysterious Singularity University, and we said that we’d positively impact the lives of a billion people in a decade with our AI Researcher.
We were, and continue to be, driven by impact
While we have not ourselves positively impacted a billion people just yet, the technologies we are building, building with and building for, have. Our insistence on many factors; most notably insisting on working on the most complicated text understanding there is (science and technology) with little to no room for errors and high requirements of privacy, and insisting on building product and not selling consultancy hours was born from this. From our very first retrieval system for research papers, to our extraction tool for accurate data systematizing, we’ve always required that what we do matter, and be solid.
Publishing research as a chronically underfunded company
We have on several occasions received a very logical question: Why the insistence on researching and publishing, when it would make more commercial sense to just try and scale the solutions you do have? While the argument makes sense, we have always known the general speed and direction of the AI space. We have held a (perhaps contrarian) belief that having a company where solid research and understanding of the field will be the right moat in the long run. From our first paper on our own WISDM metric, to our latest on our ConSens metric. (Our obsession with metrics and the ability to measure the output of language models could be a chapter on its own)
It is all about the people
Three out of four co-founders are still active in Iris.ai. We’ve stood by each other's side through ups and downs, of the company and our personal lives. We’ve been through sickness and health, death and new life, marriages and divorces, near-bankruptcies and major funding events. There’s no one else I’d be on this journey with, for both the past decade and the next. And while we have of course had a lot of great people join us, and eventually leave for new adventures - some stay. Our very first employee Vova is just a couple of months away from celebrating his own ten year anniversary with us. We must be doing something right, with the quality of people who want to work with us.
The hard work coming together in a new packaging
There are moments in life where you see the dots connecting, the threads of the past coming together. Our insistence on not selling consultancy hours, but shipping products, getting our solutions into scalable production. Our decade of research into retrieval strategies and evaluation metrics. Our incredibly hard work in building a scalable system for extracting data. Over the last year, we’ve done a substantial commercial pivot where we have been able to put everything we’re really good at into a solid developer solution - and this has also placed us in a very unique position commercially.
What’s next?
We are super excited - as a founder team, as a full team and with our investors - about the position we are taking in the market. We get to work with some incredible clients on the full scale of their AI strategy, help them cross the barriers of scaling up from PoC to production and achieve measurably good results, with tangible ROIs. We absolutely love it when parts of that implementation includes working with R&D and their research needs. We believe in privacy, security and accuracy and the power solid AI implementations have to positively impact our clients, their employees and the world.
Simultaneously, we are still thinking in ten year perspectives, in this rapidly changing world we operate in. We want to continue to provide that positive impact we’ve always been focused on. While we’re not quite ready to do a TEDx about it yet, we have some beliefs - some might be a tad contrarian - about the direction the world and the technology is heading in, and what will be our moats and unique skills in that world.
We can’t wait to see what the next decade holds!