Data is the key ingredient of modern AI, but people are increasingly wary of sharing data about their personality and behaviour with model builders.
A new architecture from researchers at the Allen Institute for AI called FlexOlmo could present a potential workaround for solving this problem.
The researchers say their approach could be particularly useful for applications involving sensitive data, such as information regarding health issues.
The technique might be cumbersome for model developers who are focused on performance, but it could be a powerful way to open up datasets that have been locked away due to security or privacy concerns.
MORE