Why AI wants to be open

The case for open AI 

There is little doubt that we are in the middle of a surge in Artificial Intelligence research and development. New advances are being aggressively pursued on a wide variety of fields from autonomous driving to automatic scheduling, to name a couple.

One of the aspects that we spend more time reflecting on, at Iris.AI, is the one to do with who will capture the expected benefits from this ongoing technological progress. Or, to be more precise, what shape will the who-will-drive-it and who-will-reap-the-benefits equation take?

In the good ol’ days the answer to this type of question was relatively straightforward. Prospective inventors understood the intellectual property legal framework well enough to be able to figure out where they stood. They had to race to a finish line, largely leveraging their individual effort, and then a number of more or less predictable things would trigger.

That traditional IPR system has come under growing criticism over time, but as professor James Robinson –co-author of the highly refreshing and thought provoking essay ‘Why Nations Fail: The Origins of Power, Prosperity, and Poverty’– reiterated at a recent conference, that patent-based system could be categorized as an inclusive economic institution largely beneficial to society as a whole. It had the merit of aligning effort and reward.

Fast forward now to 2016 and you will find a digital environment where there is ever less clarity over what will the equivalent system look like in the future. And in addition to this, a lot of very smart people happen to be expressing serious concerns about where this fascinating AI road might lead us to in the coming years. The highly publicized launch of Open AI is but a testimony to those concerns prompting action — and not an insignificant one!

One of the key things that have changed from then to now, we would contend, is that merit has become fuzzier and more complex. Leaving aside some sordid stories about ruthless appropriation of other colleagues’ work, the basic rule that held true in the past was that merit accrued to the inventor registering the patent.

When it comes to AI –and supervised machine learning in particular–, however, a new category of stakeholders is emerging: AI trainers. How will their role be recognized in distributing the value generated through future incremental algorithm advances?

These issues might seem a bit far-fetched today, but we believe they provide a solid foundation for the vision we have chosen to embrace: one of developing our AI in the open, with maximum levels of transparency and encouraging broad reusability to address different use cases that should bring about different social and economic benefits.

That is our commitment at Iris.AI, both to our staunch and growing supporter base and to the community at large: one of transparency and openness every step of the way, whilst we pursue our ambitious product roadmap going forward.

We hope we can contribute to a healthy discussion and, hopefully, that more people join us in promoting this vision. One thing we know for sure: we will not get very far unless we are joined by a critical mass of researchers, AI practitioners, corporate innovators, fellow startups, investors, advocates and knowledge seekers.