Eviden leaders discuss the power of AI

ISJ hears from Eviden leaders Dr. Boghos Boghossian, CTO, Computer Vision, Gauvain Girault, CTO, Advanced Computing and Zeina Zakhour, CTO, Cybersecurity.

We are hearing a lot of things about the transformative powers of AI, as well as the risks and challenges involved.

But the best way to really grasp this is to speak to those who are ahead of the curve in their experiences utilising AI and dealing with its challenges.  

Having said this, there isn’t a better starting point than Eviden, given its wide range of state of the art AI-powered solutions.

So, let’s make sense of AI, beyond the slogan, through insights of three esteemed Eviden leaders: Dr. Boghos Boghossian, CTO of Computer Vision; Gauvain Girault, CTO of Advanced Computing and Zeina Zakhour, CTO of Cybersecurity. 

Dr. Boghos Boghossian 

Surprisingly, all three of them seemed to be more occupied by the challenges that have emerged with the advanced use of AI.

When asked about AI potential in computer vision, Dr. Boghossian counter-intuitively answered: “The time of proving AI capabilities has actually gone!

“We have already witnessed how, where there is data, AI will learn and find patterns that outperform human counterparts.”  

For Dr. Boghossian, the focus should be on building trust in this technology, as it becomes exponentially more defining in our everyday lives.

This is not just the job of those utilising AI in their solutions. It is the responsibility of society at large.

He added: “Governments are working on making AI safe and unbiased. Scientists are working on making AI transparent and reliable.

“Trust in AI is a difficult goal to achieve, especially with its extraordinary capabilities.”  

Dr. Boghossian is also mindful of the legal challenges which will influence AI utilisation.

The legal system has evolved to help lead advanced societies – now, as Dr. Boghossian added, “these societies are working towards the next phase of evolution where self-driving vehicles are meant to operate with the boundaries of laws!

“Language models are providing useful advice and assistance and not used for cheating in academic exams. And AI is constrained to be a friend and not a foe.”  

“We will continue to ensure that AI is a core element but is surely not taking the lead.

“The importance of human intelligence and intervention to build the solutions for the world we live in is of paramount importance.

“This philosophy has to be reflected in how we develop the right solutions and how we deploy the right tools whilst utilising AI where needed.” 

The more we spoke to these three CTOs about AI, the more we realised that any serious and long term utilisation of AI has to be as invested in its challenges as it is in its potential.

As Dr. Boghossian put it: “We [Eviden] provide solutions. We therefore make sure that as part of these solutions we build trust in the tools and products we provide, including the extraordinary AI capabilities we have mastered and perfected.” 

Gauvain Girault 

Dr. Boghossian’s thoughts resonated with those of Edge AI CTO, Girault.

The challenge here, as Girault put it, is to build security infrastructure that does not impede the potential of Edge AI, given that “its servers operate outside the perimeter of traditional data centres”.

Girault explained how Edge servers need to be designed with “built-in security features such as root of trust, encryption, access control and intrusion detection/prevention to protect against cyber-threats”.  

This is only one out of a handful of differences that Girault and his team are dealing with and innovating around traditional servers and edge servers.

He explained to us, for instance, the need to process data in real time, making low latency a critical requirement.

This is particularly important for use cases such as quality inspection and maintenance decision support, where delays can lead to production downtime or incorrect decisions.  

As an example, in server manufacturing, quality inspection is critical to ensuring that products meet required standards.

One use case that Girault’s team has implemented is conformity checking of cold plates assembling.

With edge AI, organisations can deploy computer vision algorithms in the factory on the assembly lines to inspect and verify whether the cold plates are assembled correctly.

This not only improves quality but also reduces inspection time and labour costs.

He also addressed ruggedisation of servers deployed in harsh environments, energy efficiency, cost-effectiveness and proper remote management interfaces. 

Zeina Zakhour

CTO Zeina Zakhour thoughtfully highlighted the double edge sword of generative AI: “It emerges as a powerful ally that bolsters cybersecurity defence but is also a potential battleground as it extends the attack surface and introduces a new set of vulnerabilities and challenges”. 

Zakhour is direct about the “fair share of inherent vulnerabilities” in GenAI solutions.

She referred to the Open Web Application Security Project (OWASP) identifying the top 10 LLM Risks that demand attention to effectively secure GenAI.

These risks range from prompt injections to supply chain risks.

Zakhour suggested that if we are to take risks seriously, we have to admit that they span the life-cycle of Generative AI projects, posing threats of unauthorised access, data breaches, compromised decision-making and proprietary model theft.   

But, much like her colleagues, Zakhour does not seem to be put off by these challenges.

Zakhour emphasises how GenAI enhances efficiency by enabling in-depth analysis and summarisation of vast datasets, reduces detection and response times and strengthens overall cybersecurity resilience.   

So, there is a vision for her solution too, which she laid out to us in the form of a five step approach:  

• Conducting a thorough GenAI risk assessment helps organisations identify necessary security controls to be implemented and ensures compliance with regulatory and ethical frameworks 
• Building an AI data governance framework which serves as a guiding force for secure data life-cycle management and addresses concerns related to data integrity and privacy 
• Implementing zero trust principles to control access to GenAI platforms including access to prompts and meta-prompts 
• Embed resiliency into GenAI platforms by fortifying the core engine and application security and incorporating mechanisms for transparency, explainability and secure coding practices 
• Implementing a continuous, proactive monitoring solution for vigilant surveillance, real time analysis and rapid response mechanisms to address and neutralise emerging threats

As AI continues to evolve, Eviden’s CTOs provided a reassuring take on its utility.

Despite the range of solutions on offer, they share a thoughtful and ethical approach to addressing the challenges and opportunities of the new digital age.

Its commitment to innovation and responsible AI development positions it as a trusted partner for businesses seeking to harness the power of AI to achieve goals.

The company is committed to sustainability, designing solutions to minimise environmental impact.