Conversation

Notices

1. Iridium Zeppelin (bananarama@mstdn.social)'s status on Sunday, 05-Mar-2023 05:08:49 JST Iridium Zeppelin

   in reply to

   • Kevin Marks

   @KevinMarks I don't think a lot of people understand that AI is necessarily only as good as your objective. We still can't rigorously define many kinds of objective. There are issues with misaligned objectives as well. Deep learning in particular is very good at doing certain kinds of feature extraction and interpolating on some kinds of mathematical models. We are starting to determine what kinds of problems these tools are useful for. Policymakers need to know this.

   • Kevin Marks (kevinmarks@xoxo.zone)'s status on Sunday, 05-Mar-2023 05:08:51 JST Kevin Marks

     The Fallacy of AI Functionality
     “Deployed AI systems often do not work. They can be constructed haphazardly, deployed indiscriminately, and promoted deceptively. However, despite this reality, scholars, the press, & policymakers pay too little attention to functionality. This leads to technical & policy solutions focused on “ethical” or value-aligned deployments, often skipping over the prior question of whether a given system functions, or provides any benefits at all.” https://dl.acm.org/doi/fullHtml/10.1145/3531146.3533158

     Adrian Cochrane repeated this.