top of page
  • Writer's pictureBenjamin

Discovering Max-ICS and its strategy with Mr Franz Fayot Luxembourg Minister of Economy


EarthLab Luxembourg S.A. had the honour to welcome the minister of the economy, M. Franz Fayot, to celebrate its FEDIL Innovation Price received in the digital category for its Max-ICS Platform. Together with his delegation, he received an introduction to the EarthLab’s activities in relation with the Max-ICS platform in the fields of innovation, technology, and artificial intelligence. The official visit gave EarthLab the opportunity to present its technological strategy as well as its ethical vision and its values with respect to AI, data protection, cloud services and Green AI.

EarthLab Luxembourg S.A. had the honour to welcome the Minister of the Economy, Franz Fayot, with a delegation composed of Mario Grotz, representing the directorate for infrastructure and technology, Marc Serres head of the Luxembourg Space Agency, and Florence Ensch as representative of the directorate for Development Cooperation and Humanitarian Affairs. The official visit took place in the context of EarthLab's innovation award in the digital category by the FEDIL. The prize was allocated for the outstanding efforts in research and innovation which is even more critical in the current context of the sanitary crisis, as highlighted by the Minister of Economy, Franz Fayot, during the award ceremony in December 2020.

The fundamental ambition of the presentation was to foster the importance of innovation and its vast potential, specifically in the field of Artificial Intelligence (AI). The visit was organized in two parts. First, the visitors were familiarized with the company’s history and offered products and services. Hereafter, a practical demonstration through concrete use cases was organized.

EarthLab’s capabilities rely on the Max-ICS platform, the teams’ undeniable devotion and know-how, and their unique perception of values on AI translated in the platform. The Max-ICS platform has been designed to implement digital applications, treatment chains, and end-to-end solutions easily to create machine and deep learning models. The platform offers a zero-code based catalogue of technologies and innovative tools.

Examples of vertical applications

Over the past five years of activity, EarthLab considerably extended its domain of activity. The former specialist in risk assessment in the insurance ecosystem has enlarged its expertise by accepting new challenges and seeking cutting-edge technologies and innovation. To date, EarthLab counts the following domains to its vertical products developed thanks to Max-ICS: agriculture, maritime surveillance, humanitarian actions, fraud and cybersecurity, business forecasting, space surveillance or audio-visual archiving.

And beyond…

The company's success relies on its strategy to address AI technologies within digital applications and big-data treatments. EarthLab pursues its innovation strategy and will work on four pillars during the next period, continuing to develop its vision for a sustainable AI: (1) valorise AI models as an internal company asset, (2) ensure the safety of AI models, (3) guarantee a certain degree of explainability and interpretability for the delivered models and predictions and finally (4) promote a green, environmental-friendly AI.

Giving secrecy and value to generated models

First, it is important to recognize that AI models represent an important value for the company. On the same scheme than software or datasets, AI models should be perceived as an internal company asset and thus must be protected accordingly. The Max-ICS platform solution is a secure product, offering end-to-end security and data as well as model protection management. To guarantee data privacy, security, and protection of the platform’s users, Max-ICS can rely on different cloud solutions, adaptable to partners and projects requirements. The cloud service ranges from public clouds, where computing is placed on public cloud providers such as AWS, Google Cloud Platform or Azure. This solution is more adapted to non-sensible open-data. The set of community clouds provide an intermediate cloud system offering with an increased data privacy. Here, the computing is more isolated and thus adapted for commercial and confidential data. The most protected formula is the on-premises deployment designed for totally isolated and personal data for which the data and the models are secured within the source company infrastructure. This range of deployment solutions can be used for any aspect of Max-ICS and ensure the privacy of the generated models including their application. As an extension, EarthLab is actively working on data-privacy innovation introducing in Max-ICS new concepts such as Differential Privacy, Encrypted Computation and Distributed Computation.

Safety aspects of executing AI models

Second, it is important to consider the safety of the model execution (so called inference) to ensure the high-quality of AI model results. Hence, the model outputs must be analysed and checked to depict potential hyperactivation inducing misleading or duped model results. Such hyperactivations can be caused by external effects or manipulations, such as adversarial patches implemented to fool machine learning models by positioning physical obstruction elements in captured pictures.

Explain and interpret the AI to unlock the black box

Third, AI explainability and interpretability are necessary to explain and interpret model outputs. Indeed, artificial intelligence systems often leak transparency leading to the well-known black-box effect perceived as a barrier to adopting complex machine learning models. Max-ICS includes explainability and interpretability methods to foster a user-friendly utilization of complex machine learning models: while it does not provide a way to understand how AI works, it contributes to the understanding of its comportment. Besides, providing strategies to facilitate model outputs’ description, interpretation and reliability, Max-ICS additionally supports model training traceability.

Red and Green AI at your hand

Last, besides numerical and digital transformation, the environmental and energy transition is a main priority of the current government. Contrary to the general opinion on AI’s environmental impact, training a model requires 10 to 1500 experimental runs which lead to a consumption of several tons of CO2. Thus, the trade-off between accuracy and energy consumption become an important consideration and open the path to select between Green AI, moderately less accurate but less consuming, or Red AI centered exclusively on accuracy. Green AI is still a new area in which, through the Max-ICS platform, EarthLab is conducting initial Research & Development activities to provide user-friendly solutions to limit the environmental impact of AI.

117 views0 comments


bottom of page