AI Redefined (AIR)
Putting people at the heart of AI
Putting people at the heart of AI
AI is a revolutionary advancement and it is now time for humans to take their place at the core of these technologies: this is the mission of AI Redefined.
AI is the latest trendy technology. Seen by many as the champion of hope for solving our environmental, economic and social problems, it also raises concerns related to employment, control, and privacy. Whatever one’s position is on these issues, one thing is clear: AI is transformative and we’re only at the beginning.
Dorian Kieken and Fabrice Condominas, AIR co-founders.
For AIR, the challenge of AI democratization revolves around people. Even though machine learning’s technical progress is impressive, it essentially remains the progress of a dehumanized AI, used as an advanced tool in narrow applicative contexts. Without delving into the general AI dreamed in sci-fi works, AIR is already working on what’s coming afterwards: how this new form of cognition can harmoniously be weaved into the fabric of our societies and how to support, complement, and elevate human cognition. In other words: how to develop artificial agents that go beyond just being tools. How can we help humans and societies to think better and face challenges already upon them in an age that is increasingly complex and accelerating towards civilizational difficulties we feel overwhelmed by — from climate change to energy production, to more equitable and sustainable use of our resources, material and otherwise, to dealing with our waste. The answers to these questions undoubtedly lie within a more holistic approach, and with the development of tools that will allow AIs to evolve alongside dynamic and still continuously learning humans. These AIs need to not only learn from our data, but also from the way we interact with our environment, our peers, and other living beings on the planet. Evidently, the first step towards this vision calls for the development of new tools, technologies and approaches that will allow both AIs and humans to learn together, continuously, and at scale. We’ve named this technological framework – Cogment.
AIs of the future need to develop in conjunction with humans, in all their plurality and diversity.
(Proof of concept) Developed on AIR’s framework for the city of Hong-Kong, a Recommender agent helps first responders’ services operators to tackle the challenges of large urban centers.
For AIR, the challenge of AI democratization revolves around people.
Intelligence is a complex concept. It exists in many forms. It’s undoubtedly a tool for innovating that we’ve been using for millenia. If we are to keep innovating freely with it, it therefore seems important to rid ourselves of methodological, architectural, or scale constraints, as we’re starting to explore its artificial incarnation. It’s with this idea in mind that we created Cogment. Cogment is the first open source framework dedicated to the continuous training of multiple agents alongside multiple human users in shared environments. Designed to be evolutive and technologically agnostic, Cogment doesn’t enforce any particular machine learning approach. Supervised, unsupervised, reinforcement, imitation, curriculum, or genetic programming; all approaches from the past, present, and even the future, work with Cogment. Cogment is also compatible with existing tools, whether they are machine learning libraries employed in agents themselves, such as Tensorflow, Keras or Pytorch, or technologies for deployment and communication, such as Kubernetes or Docker. Cogment enables new paths to innovation which light the way toward a distributed concept of AI design in which complex architectures are now possible. Since Cogment is developed as a free open source framework, researchers, academics, and even hobbyists can use it. What marvelous innovations will the AI community come up with using this new tool? Canada, Québec, and Montreal, are already recognized as world-renowned epicenters of artificial intelligence; by developing its humanist and ambitious vision for AI locally, AIR is not only availing itself of this talent pool, but also allowing the local community to benefit from it. This is already the case with Thales, for example, who is using Cogment on several projects after having partnered with AIR during the [email protected] program in 2019. Although the first application areas of this approach, like defence and aerospace, are somewhat removed from people’s daily life, AIR developed this technology in open source with the singular goal of allowing it, in due time, to serve closer domains that the company already tackles; supply chain, urban planning and smart cities, education, healthcare, video games to name but a few.
Grégory Szriftgiser, co-founder
François Chabot, co-founder
Whether we want to elevate our decision, prediction, analytic, teaching or communication capabilities, tomorrow’s Artificial Intelligence needs to evolve alongside humans, in all their plurality and diversity. If we’re to tackle tomorrow’s challenges, to better make sense of our reality, and to move forward together, we must create, support, and popularize this continuous, bi-directional learning loop between artificial intelligence and our own. Beyond our own performance, we will also no doubt find numerous opportunities to free up new potential and focus our effort on what is truly important while delegating the remainder to our new allies.