MIT Internet Policy Research Initiative

AI Policy – 2024

IPRI’s AI Policy works focuses on trustworthiness, transparency, security, and privacy related to artificial intelligence (AI) and machine learning (ML) systems. 

IPRI researchers are at the forefront of developing tools that allow for secure machine learning over encrypted data. These tools can work well for well-understood tasks where the hyperparameters of the training can be set at the start. For example, taking a public model and then fine-tuning it on proprietary data.

We also focus on autonomous vehicles. If we want users to trust the cars they are in and feel comfortable and willing to relinquish control, autonomous subsystems will have to be able to explain why they took certain actions and show that they can be accountable for errors. Explanations will have to be simple enough for users to understand, even when subject to cognitive distractions. In response to this, our machine understanding work explores techniques for enabling autonomous systems to explain themselves by generating coherent symbolic representations of the relevant antecedents of significant events in the course of driving.

 

People – AI Policy

People – Alumni – AI Policy

Research and Events