Sign up


Evo Pricing chosen by Facebook in the task force to reduce the risks of Artificial Intelligence

MILAN – Evo Pricing, a company specialized in artificial intelligence with offices in London and Turin, participates in the Facebook AI Open Loop program. Incubated at the Polytechnic of Turin, Evo is one of ten companies chosen by Mark Zuckerberg’s company to lead the European task force charged with defining a governance to reduce the risks of artificial intelligence (AI). The only other Italian is the startup RiAtlas, specialized in e-health.

With public opinion divided on the impact of AI in society, many fear that unregulated development poses too great a risk. The European Union has been dealing with the issue and its legislation for some time. Open Loop aims to create a new regulatory framework with forward-looking, fact-based guidelines that can effectively identify and manage AI risks while encouraging innovation. The program supports EU policymakers in finding practical solutions that minimize risks to the public, without putting European technology companies at a competitive disadvantage due to overly burdensome AI governance and, in ultimately, ineffective.

The contribution of Evo Pricing alongside Facebook

Evo Pricing has chosen to contribute by engaging more actively in the development of AI legislation to find policies that protect the privacy of individuals, while using their data effectively to better meet their needs. “The Open Loop program launched by Facebook has fostered a healthy debate within Evo Pricing on future AI governance issues that span a period of 3-5 years, beyond our typical planning horizon of 12 months. Governance is a fundamental issue for us at Evo Pricing, as we collect and process data on over 1.3 billion people and over 900 million products globally every day” said Fabrizio Fantini, founder and CEO of Evo Pricing.

The resulting framework, called Automated Decision Impact Assessment (ADIA), designed in a very similar way to the GDPR Data Protection Impact Assessment (DPIA) framework, has proven to be very promising in achieving these goals.

The satisfaction of Facebook and the first results

“The experiences of our partners have shown how this type of risk assessment approach can promote a more flexible, practicable and innovative method for assessing and managing AI risks than more prescriptive policy approaches,” said Norberto Andrade, Global Facebook AI Policy Lead for Digital and AI Ethics.

The program itself used a cutting-edge approach known as policy prototyping to create its AI governance framework. An empirical approach to experimental governance during which, Evo tested the AI ​​policies proposed in its application of Autonomous Supply Chain to measure the impact that the framework has had on operations, AI innovation and risk management.

The resulting data provided Evo with feedback on the impact, strengths and limitations of the framework. Areas of confusion or compliance challenge could be identified early to iterate and improve these regulatory frameworks. In this way, the Evo team can now provide policy makers with practical input to improve existing regulatory frameworks and update legislative processes.

“This experience prompted us to think about how to prototype a political document that makes our autonomous supply chain fit for the future and at the same time further promotes the agenda and public debate,” said Fantini.

Evo and the partners of the Open Loop program are looking to consolidate ADIA’s early successes to further increase transparency and accountability.

“Policy prototyping has helped my team discover new areas where we could include stakeholder feedback and transparency. We have already increased manually editable parameters by 16%, to give our customers more direct control, and set stricter limits on the solutions we generate. Based on simulations, this will make our algorithm even more accurate. More innovative governance is a win for everyone, and we have high hopes for future iterations”, concluded Fantini.

Hey! Was this page helpful?

We’re always looking to make our docs better, please let us know if you have any
suggestions or advice about what’s working and what’s not!

Send Feedback