Artificial intelligence (AI) is becoming everyone’s business as use cases expand across many industries. However, really understanding what is behind the algorithms and what this means for business is no small feat.
The proof is, according to a recent BCG study: while 84% of companies say responsible AI should be a priority, only 16% of them have developed a mature program.
To overcome these difficulties, the cosmetics leader L’Oréal, the operator Orange France, the consulting company BCG Gamma and the insurer Malakoff Humanis are launching a responsible AI label. It is designed to help companies spot failing algorithms and get rid of bad practices.
Risky algorithms
Presented this Thursday to the press, this label should be set up “early 2023”, indicated Laetitia Orsini Sharps, president of the Positive AI association, carrier of the project. It will be awarded to players who meet the criteria defined by a reference system created with the support of EY.
The label is in line with the future European framework on responsible AI, which must be applied from 2025. Companies applying for the label will have a panel of their “most relevant” algorithms tested by an independent audit firm. risk “. These audits will be repeated at regular intervals, to prove the transparency and robustness of AI systems over time.
Each in their sector, the four companies behind the label have been confronted with risky algorithms. To limit or even eliminate the risks, they carried out an in-depth analysis of the possible biases of the training datasets, of the governance, of the life cycle of the algorithms and of the ability of the tech teams to open the hood and explain how it works.
Embark the business
The objective, with this label, is to “get businesses on board” in responsible AI, these players being the main ones concerned by the development and deployment of AI tools on a more or less large scale.
Responsible AI is not only used to limit risks, it also serves a business interest. “We often assume that companies that embark on responsible AI want above all to avoid risks, but it is also to create value,” says Nicolas de Bellefonds, global director of BCG Gamma.
During 2023, Positive AI wants to set up a digital platform for companies to carry out self-diagnoses on their level of maturity.
Beyond concrete tools, Positive AI sees itself as a space for discussion between AI experts and top managers. According to Laetitia Orsini Sharps, the vocation of the association is to open up to more companies from different sectors.