To overcome two challenges in training AI – scarce or hard-to-get data and data privacy – researchers have come up with a counterintuitive technique: fake it.
Articles on Training data
Displaying all articles
Once errors creep into the AI knowledge base, they can be very hard to get out.
As DOGE taps into sensitive federal agency data repositories, many people fear what could happen to the data. One little-discussed but hugely consequential possibility: fueling Elon Musk’s xAI company.
Many videos people upload to YouTube aren’t really meant for public consumption, but they’re available for AI companies to vacuum up. Many of these personal videos are posted by children.
Bias in AI has been getting a lot of attention lately, but it’s just one aspect of the larger – and thornier – problem of fairness in AI.
When it comes to training high-performing AI models, the quality of the data is just as important as the quantity.
Top contributors
-
Associate Professor of Public Policy, Communication, and Information, UMass Amherst
-
Associate Professor in Law, Macquarie University
-
Senior Research Fellow, Initiative for Digital Public Infrastructure, UMass Amherst
-
Assistant Professor of Computer Science, University of Virginia
-
Distinguished Endowed Professor, Middlebury College
-
Research Fellow, GenAI Lab, Queensland University of Technology
-
Senior Research Fellow in AI Accountability, Queensland University of Technology
-
PhD Candidate, GenAI Lab, Queensland University of Technology
-
Professor of Statistics and Professor of EECS, University of Michigan