ACE, Artificial Intelligence made in Safran for defense equipment
At the World Defense Show 2024 in Riyadh, Safran Electronics & Defense unveils ACE, Advanced Cognitive Engine. Its objective: to integrate AI capabilities into all our products, to provide operational staff in the field with a better understanding of the situational environment, whatever its complexity.
Safran Electronics & Defense's optronic and navigation systems can now benefit from an "Artificial Intelligence" function: ACE. The aim is to meet the operational needs of Forces in the field by lightening their cognitive load. Indeed, automating complex, repetitive actions enables operational staff to refocus on high value-added tasks.
"An ACE" is a person with advanced knowledge and skills in a complex subject or field. With ACE, Safran Electronics & Defense has taken a new step forward, putting onboard intelligence at the service of the human intelligence of Forces in the field. Intelligence On Board, ACE On Board!
How does it work?
In the field, thanks to ACE, operators have a better understanding of their environment.
On-board intelligence, based on data learning via neural networks (deep learning), opens up a whole range of new capabilities. For example, it enhances tracking, detection, classification and target identification capabilities. But it can also correct the effects generated by environmental conditions (atmospheric turbulence, low light levels). ACE will also enable Forces to supplement their own databases and thus improve the on-board learning of their equipment.
Find out more about the role of AI in the military sector
Already today, and even more so in the near future, AI has a decisive role to play with the Armed Forces in the field. What role? What specific challenges must it meet?
Three questions to Estelle Parra, AI Program Manager in R&T at Safran Electronics & Defense.
Is the integration of AI into our defense equipment something new?
Yes... and no! Without getting into semantic debates, putting intelligence into products, as defined by NATO, is nothing new. At Safran Electronics & Defense, we've been using computer vision in our seekers, airborne optronic systems and land-based sights for a very long time: image enhancement codes and designation codes, for example, are at the heart of our defense equipment. We use the term "Intelligence" to designate all these computer-coded applications.
The notion of Artificial Intelligence became increasingly popular in recent years. It stems from a dual technological movement:
- computer coding capabilities, through complex architectures
- and the possibility of integrating these mathematical developments (neural networks or deep learning) thanks to the evolution of microprocessors (CPU-GPU).
It's the meeting of electronics and mathematics: boosted computers and multiplied coding capacities make it possible to entrust the machine with all or part of the choices previously made by man.
These far-reaching changes mean that we can now work concretely on subjects that we thought impossible just a few years ago!
What are the major challenges of AI for the defense sector?
Thanks to fantastic improvements in electronics, we are able to develop ever lighter and more powerful AI systems... where 10 years ago, to operate an autonomous vehicle, you had to fill your trunk with computers!
However, the military sector is, quite rightly, absolutely demanding in terms of the reliability of the information provided by AI. One of the main challenges is to put in place a system capable of controlling variations in calculations, so as to guarantee confidence in the results provided. This system of confidence, essential for the defense sector more than any other, is at the heart of the Confiance.ai program, which brings together the scientific community involved in the subject. Safran Electronics & Defense is working with the other members of the program to develop common functions and processes designed to guarantee this trust.
A second challenge we face in expanding the use of AI in the defense sector is that of transferability, or transfer learning: how can we guarantee that what AI has learned in one particular field (known as a field of use) will be just as efficient in another? This is a critical issue for the Armed Forces: deployed yesterday in the Sahel and tomorrow elsewhere, how can we guarantee that AI will provide the same capability function in a very different environment?
How can we guarantee the armed forces that our AI-enhanced products will deliver the added value they need?
We need to know exactly how the end user - in this case, the Armed Forces - uses our equipment. High-performance AI requires fine-tuned interaction with humans (Human in to over the Loop): it is therefore imperative to understand how to influence the decision that humans will ultimately make. Working hand in hand with users must help us to meet their needs... even if this means changing our way of thinking, and the concepts and paradigms we thought were immutable!