•  
  •  

Madrid, November 2022

eProsima is happy to announce the approval of the R&D project SustainML as an EU funded AI project.

SustainML will be coordinated by eProsima, and the main objective behind this project is to achieve a significant reduction in the CO2 footprint of ML applications. To accomplish this, energy-aware applications must be as easy to develop as standard ML systems are today. 

To achieve this goal SustainML aims to develop a design Machine Learning framework and an associated toolkit for Green AI that will foster energy efficiency throughout the whole life-cycle of ML learning applications. This way users with little or no understanding of the tradeoffs between different architecture choices and energy footprint should be able to easily reduce the power consumption of their applications.

Objectives

In order to achieve SustainML's mission to reduce the CO2 footprint of ML applications, a number of objectives have been defined:

O1: Model the requirements of specific ML applications. A new paradigm to break down ML tasks into structures of elements that represents the taxonomy of the vast majority of ML applications

O2: Resource aware optimization methods based on models from previous objectives. A toolkit that can map requirement specifications ( from O1) onto different estimations. 

O3: Footprint and AI-waste transparent interactive design assistant that guides the developers through the entire process. A software tool in the form of an interactive wizard that guides the developers through the entire development and implementation process, to help them achieve their design goals while minimizing the energy consumption.

O4: Collection of efficient methods and cores as catalogs and libraries of energy optimized parameterized ML models. A library of parameterized and energy-optimized ML models to implement the most energy-efficient applications possible. 

O5: Dedicated toolchain implementation. A tool chain dedicated to actual implementation, testing and validation based on the above components. 

 

Expected impact

The SustainML framework will address the carbon and resource footprints of ML models and offer multiple pathways to avoid AI-waste from the very early stages of AI life-cycles. This will not be a limiting factor for the rapid growth of both AI research and AI adoption, but rather an enabling tool focused on sustainable growth.

IMPACT 1: On a global level reducing AI waste through the entire life cycles to a wide spectrum of application domains targeting all AI developers from novice to experts. 

IMPACT 2: While AI will not solve all problems by itself, the ability to greatly expand AI into all industry sectors and use its means of optimizing resource consumption more widely, without AI itself being a major CO2 emission and financial cost factor, will be hugely impactful.

IMPACT 3: Transition from the “democratization of AI” to the “democratization of Green AI” that will allow especially SME, private enthusiasts, NGOs and individual innovators to develop and use AI in a sustainable way.

IMPACT 4: Cultivating trust in AI for the general public, and insight in AI for non-expert practitioners by systematic, transparent and explainable taxonomy and knowledge recycling approaches.

IMPACT 5: Being a nexus to spread new technologies/hardware such as PIM that can offer a sustainable alternative to current HW.

 

The SustainML consortium is a mix of partners includingbig corporations, small and medium-sized enterprises , academic institutions and EU-wide acting foundations with backgrounds in AI, robotics, and general technology. The members are: DFKI, IBM, UPMEM, Inria, University of Copenhagen, Technische Universität Kaiserslauternand eProsima.

 

FOR MORE INFORMATION ABOUT SUSTAINML:

Please contact This email address is being protected from spambots. You need JavaScript enabled to view it..