In our conversation with Fiona, we dive into the importance of data quality. Fiona is a strong advocate in applying rigour to the data process in machine learning. She firmly believes that the unglamorous side of machine learning needs to be addressed and standardised to avoid unethical issues related to these technologies.
Interview with Fiona Browne, Head of AI at Datactics. You can follow Fiona on LinkedIn.
Recently, there has been much more awareness of the importance of data quality. In our interview, Fiona spoke about Google’s recent paper, emphasising a rigorous data process. Fiona explained that in the creation of machine learning, there is more focus on building a model rather than the data used to power it — the data side is not as glamourise as building a model.
However, Fiona emphasised the importance of data quality. It would improve the performance of the models and have a better impact on society. When it comes to high-risk AI, data quality has a significant effect from the get-go. For example, with GTP3, the voice recognition model was racist and biased against minority groups. A data governance framework is needed to ensure that these issues are picked up during the creation process and not when interacting with customers and other members of society.
When building a data governance framework, it is vital to have a diverse range of voices. Fiona also highlighted the importance of having a devil’s advocate group to ask different questions to ensure that all bases have been covered.
What are your thoughts? Join our Slack channel and join the conversation!