Simha Sadasiva CEO and co-founder Ushur recently spoke to a professional that mentioned a surprising anomaly. Their newly purchased laptop, undoubtedly sleeker and more powerful than their previous one, had a flawed spell check function.
Sadasiva recently explored how data integrity is changing AI.
For many over 30, spell check represented their initial interaction with automation. Intriguingly, while most technologies, like cameras in iPhones or car fuel efficiency, have seen continual advancements, why did such a basic feature like spell check seem to regress?
The answer is inextricably linked to data. The modern world is saturated with information. One would assume that the profusion of data should equate to smarter decisions and enhanced outcomes, it said.
However, AI’s efficacy is directly proportional to the quality of data it’s trained on. Early spell check systems trained on rigorously edited published texts. But now, with the internet providing everyone a publishing platform, and stringent editorial standards being relaxed for quicker news delivery, there’s a surge in content, not necessarily maintaining the same quality. Hence, even basic software like spell check isn’t as efficient as it once was.
Data integrity isn’t just about spell check, it’s monumental. On one hand, it pertains to ensuring user data remains confidential, safeguarded against misuse. Any modern software provider is anticipated to maintain these standards. On the other, data integrity is pivotal for actionable insights. For instance, a runner who had used Garmin watches for monitoring vitals for years faced trust issues when a newer model showcased inaccurate heart rate readings due to a software glitch. This inaccuracy had real-world consequences, shaking loyalty towards a brand they had trusted for nearly two decades.
Changes in data structuring have also emerged as a pressing concern. Whereas earlier data analysis was more systematic, the proliferation of unstructured data now presents anomalies in data interpretation. One such instance is a Tesla owner who felt recent software updates compromised the car’s Full Self-Driving experience.
Still, the magnitude of data available today is unparalleled. Devices, from phones to IoT gadgets, generate an overwhelming amount of information. The challenge is discerning relevant insights. Without this capability, data loses its significance.
At InsurTech company Ushur, the significance of upholding trust is observed daily. Especially in sectors like finance, healthcare, and insurance, professionals are reticent to adopt technologies that might introduce even a whiff of risk. For instance, an AI reading a CT scan must be flawless; even a 95% accuracy rate is unacceptable. However, if machine training models are set up meticulously, their potential is transformative.
Ultimately, leveraging data, its analysis, presentation, and subsequent actions offer monumental prospects. Yet, if executed poorly, the implications can be calamitous. Therefore, AI’s next innovative phase will focus on eradicating biases, setting boundaries for Large Language Models (LLMs), and maintaining data security and privacy.
Read the full story here.
Keep up with all the latest FinTech news here.
Copyright © 2023 FinTech Global