The role of data, networks, trust for artificial intelligence take up

Enterprise IT News has a chat with Juniper Networks’ ASEAN Systems Engineering Director, Sui Jin Foong.

EITN: Your study surveyed trust levels towards artificial intelligence (AI). But there are challenges in deploying the technology, despite recognition of its benefits. Is it an issue to do with the lack of solutions or lack of talent/skill to deploy/operate the solutions?

Sui Jin: In the Juniper Network’s global research report, it was found that while the APAC region displays a strong level of trust and acceptance towards AI, both the lack of strong talent/skills and the lack of solutions play a role in hindering the adoption of AI.

In any deployment, it is necessary to find a suitable solution that works best with the skills and processes within the company. There exists a need for stronger infrastructure including data, cloud and networking abilities, as well as talent to work with these AI systems.

Strong infrastructure comprising data, cloud and networking capabilities – as well as the talent and skills are keys to developing effective AI models and data sets.

AI systems are complex to begin with and may require expertise to set up and operate. However, in today’s business landscape, there are many “off-the-shelf” solutions readily available in the market. These solutions can be easily adopted by companies as long as they are aware of the areas where AI can be applied with the least friction.

EITN: According to the survey, the top challenge to AI adoption includes developing AI models and data sets. How can this challenge be addressed and overcome in your opinion?

Sui Jin: According to the survey, 74% of executives are likely to collect telemetry data to enhance AI in order to enhance the user experience. Yet, respondents still rank developing AI models and collating data sets as the top technology-related challenge.

Strong infrastructure, which entails data, cloud and networking capabilities – as well as the talent to work with these AI systems are keys to developing effective AI models and data sets. The Juniper Mist solution leverages an AI model that provides simplified operational experience and optimizes end user experience within Campus Wireless and Wired networks, as well as the WAN.

When it comes to developing AI models, organizations can opt for standardised algorithms already developed for common use cases instead of reinventing the wheel. AI models can also be pre-trained with large and wide datasets. This option would serve users whose requirements are already standardised. For instance, in the space of facial recognition, there are well-known optimized models that can be used as-is.

When it comes to data sets, ensuring that unstructured data is accurate is important when training reliable and effective AI models. In the absence of reliable training data sets, unsupervised learning algorithms have been shown to be used successfully. These unsupervised learning algorithms include clustering, anomaly detection and neural networks and are seen to have found good efficacy in networking and security use-cases.

EITN: How much of a challenge is infrastructure-readiness to deploying and more adoption of AI technologies? How can it be addressed and overcome?

Sui Jin: Infrastructure readiness can be seen from two perspectives.

From the IT perspective, the use of cloud has been a big advantage for deployment and adoption of AI. With cloud-based platforms, organizations need not invest in large in-house data warehouses for machine learning. Cloud platforms also improve effectiveness since they have access to a much larger dataset across multiple organizations.

From the network perspective, there are three considerations for readiness. Firstly, the network should be able to send out large amount of telemetry, preferably real time to the AI. Next up, the network should be ready or already support a rich set of API that allows the network to be dynamically and flexibly reconfigured. Lastly, AI should be ready to ingest rich real time telemetry, allowing it to analyse, as well as to interact with the infrastructure to correct or optimize processes accordingly.

These considerations can be seen as benchmarks to ascertain infrastructure readiness. As a start, organizations should look at how they can integrate AI-driven automation into their wired and wireless network infrastructure. By shifting away from traditional wired and wireless LAN solutions that are built on antiquated architectures, organization can better deliver optimized experiences and simplified network operations.

EITN: In your opinion, why is there a higher level of trust in APAC compared to North America?

Sui Jin: Trust is defined by five elements: Privacy, Security, Reliability, Ethics, and Compliance. When a technology fares well within these parameters, it gains the trust of organizations and their customers. For example, governments in Japan, South Korea, China, India and Singapore have put in significant effort in building next generation AI capacity, laying out governance policies for AI-led initiatives.

In 2019, Singapore introduced the Model AI Governance Framework at the World Economic Forum to help promote public understanding and trust in technologies. Essentially, the framework provides guidance to enterprises to address key ethical and governance issues when deploying AI solutions.

A strong AI governance framework that encourages building good data accountability practices, maintaining open and transparent communication, and explaining how AI systems work is the way forward. This approach will not only help organizations better understand the benefits of AI-driven innovation, but also give consumers the confidence to adopt and use AI.

EITN: Your survey revealed something of a paradox: broad acknowledgement from respondents for cross-functional executive sponsorship, but yet, only 3% of executives reported they have identified an AI leader who oversees AI strategy and governance. Why is this so?

Sui Jin: This is indicative of the current state of play. While there is a strong interest in making AI mainstream, there is a gap in the talent pool.

Organizations that wish to expand into AI must often deal with the balance between hiring and training. While 48% of APAC organizations are struggling to prepare and expand their workforce to integrate AI systems, C-level respondents have reported that hiring is a priority over training employees to handle AI-driven tasks.

But there is a scarcity of executive-level AI talent who can comprehend the technical aspects of AI as well as the ramifications of governance policies for AI design.

In the absence of qualified leaders for hire, organizations are looking at alternatives. Our study shows 87% of executives in APAC agree cross-functional executive sponsorship and involvement are critical for AI to integrate into their products and services. AI adoption must be closely considered and jointly led by organization leaders and the CIO.