• Monday, December 09, 2024
businessday logo

BusinessDay

AI accuracy depends on reliable data — Experts

Artificial Intelligence models will produce inaccurate results without high-quality data, experts in the space have said. They said this at Microsoft Africa’s AI Journalist Academy, which was recently held.

Since ChatGPT, the poster child of AI models, launched in 2022, the adoption of AI has skyrocketed. This opened the door for the launch of more AI models like Gemini and Copilot. Today, ChatGPT has over 180 million users, and more use cases for AI continue to be discovered.

Dean Erasmus, Microsoft South Africa’s chief data officer, said, “The pace of innovation in AI is astonishing. The demand from customers and governments for AI capabilities is skyrocketing. Use cases that were once priorities are now standard SaaS offerings.”

Erasmus noted that most of this advancement has been possible because the falling cost of AI infrastructure has made the technology more accessible. However, as the implementation of AI grows, organisations must have the right data and infrastructure.

“Without high-quality data, AI models will produce inaccurate results. Additionally, organizations need to invest in the necessary infrastructure to support AI workloads, including powerful hardware and specialized software,” Erasmus stated.

Erasmus highlighted the importance of organisations adopting AI based on a framework that focuses on fairness, reliability, inclusivity, standards, and accountability.

According to the CDO, fairness ensures that AI systems allocate opportunities and resources equitably. “Reliability guarantees that AI systems are reliable from both a data and architectural perspective. Inclusivity is crucial to avoid excluding large segments of the population, especially in regions where English is not the primary language. Standards and practices are essential for accountability and transparency, which in turn drive innovation,” Erasmus explained.

As AI becomes more sophisticated, there is a growing risk of bias, discrimination, privacy violations, and cybersecurity threats. Wessel Pieterse, CISO of Microsoft South Africa, noted that cybersecurity threats are evolving rapidly.

Today, the speed, scale, and sophistication of cyberattacks have increased, and traditional tools are no longer enough to keep pace with cybercriminals’ threats. The threat landscape is also changing with the evolution of AI.

“Device compromises and password attacks have increased dramatically in recent years. Microsoft tracked a 10-fold increase in password attacks between 2022 and 2023. Additionally, the sophistication of threat actors has grown, with over 300 large-scale threat groups now active,” he said.

Colin Baumgart, CTO and Commercial Solutions Area Director at Microsoft South Africa stated, “The cybersecurity landscape is undergoing a profound transformation, driven by the relentless evolution of technology and the increasing sophistication of cyber threats. In this dynamic environment, artificial intelligence (AI) has emerged as a pivotal ally in the fight against cybercrime.”

The CTO noted that AI’s ability to analyse vast amounts of data at lightning speeds enables the identification of patterns and anomalies that may indicate a security breach, often before it occurs. “This proactive stance is crucial when reactive measures are no longer sufficient,” he emphasised.

However, Erasmus, earlier quoted, noted that AI’s full potential would never be realized without investment in education and training. “We need to develop a skilled workforce that can develop, deploy, and maintain AI systems. Additionally, we need to foster a culture of innovation and experimentation to encourage the development of new AI applications,” the CDO added.

Join BusinessDay whatsapp Channel, to stay up to date

Open In Whatsapp





Exit mobile version