Principal at KPMG | Global Data, Cloud, AI Leader| U.S. Data Modernization Leader | Avid Skier & Snowboarder |
As we explore the boundary between data and AI and the notion of trust in data, there is one challenge that really stands out to me. What happens to your data once you use it to optimize a large language model or enter it in a generative AI tool? In machine learning, there are limitations to removing data once it’s been used. This issue has significant implications on data retention, data contracts, and data rights and can limit how third-party data can be used. Businesses will need to consider how they manage machine UN-learning in 2025. #BigIdeas2025 #KPMGTechnology #Data #AI #DataModernization #MachineLearning
Lisa Rawls
2 个月
Insightful! Thanks Matteo - and agree!! Data governance and getting that right, while taking all the regulatory requirements and security & risk mitigation strategies into account, will continue to be critical in 2025!
Data & Analytics
2 个月
Matteo Colombo, data retention poses a serious issue. Once it's in the model, it’s tough to erase—definitely something businesses need to strategize around.
I agree Matteo!
Consumer & Retail Technology Leader
2 个月Well said Matteo Colombo. This issue amplifies as organizations are starting to truly harness and leverage first party consumer data especially around e-commerce and marketing use cases within their machine learning and AI modeling. Consumer privacy laws such as Right to Forget(RTF) are on a opt in/out basis and unless there’s a way to quickly ‘unlearn’ insights when someone opts out, we’re potentially looking at not being compliant. Having a clear strategy around data management and specifically in this case, being able to have end to end lineage and detailed and accurate data catalogs at your disposal becomes very important for our clients.