Accurate Data and AI Outcomes: The Importance of Governance and Risk Management in Next-Generation Technologies
Artificial intelligence and machine learning have revolutionized the way data is processed and analyzed. These technologies have the potential to automate tasks, make predictions, and improve decision-making processes. However, as accurate as the data used to train AI models may be, it doesn’t always guarantee accurate outcomes.
The U.K. Information Commissioner’s Office recently highlighted the importance of accuracy in AI models, especially when it comes to generative artificial intelligence. In a public consultation, the agency emphasized the need for developers to filter out inaccurate training data to ensure the statistical accuracy of the models and the correctness of personal data contained in IT systems.
The consultation suggests that developers should set clear expectations for users regarding the accuracy of the output and provide information about the statistical accuracy of the models. It also recommends retraining models based on user experiences to improve accuracy.
One key point raised in the consultation is the use of data from untrusted sources, such as social media and online forums. While these sources may have high levels of engagement, they may not always provide accurate data. This can impact the accuracy of AI models and lead to inaccurate outcomes.
AI researcher Johanna Walker from Kings College London pointed out that the outcome of a generative AI system is dependent on how users prompt the device. This means that a system can generate inaccurate outcomes even if it has been trained on accurate data.
The ICO’s approach to accuracy in AI models aligns with privacy laws that require processors to take reasonable steps to correct or erase incorrect data. This helps mitigate biases in data and ensures that personal information is handled accurately and ethically.
Overall, the ICO’s focus on accuracy in AI models is crucial for ensuring the responsible development and deployment of AI technologies. By emphasizing the importance of accurate data and monitoring model outcomes, organizations can improve the reliability and trustworthiness of their AI systems.