There’s some interesting news about Amazon Q AI this week: The AI that is supposed to give useful help for your Amazon Web Services apparently was giving more than expected…
In recent AWS documents, employees reportedly noted leaking information such as “location of Amazon Web Services (AWS) data centers, along with internal discount programs, and unreleased features”, supposedly “Amazon built Q to be more secure and private than a consumer chatbot”, but apparently this may have been trained upon data that they didn’t want to be publicly accessible?
This especially highlights the use of source material when building out a large language model. This is something that would not have been an issue with a smaller and more search-engine-ish algorithm, such as has been done in Chatterbot – with specific questions mapped to answers. Furthermore, that is would have been an easier way to build a system without massive Nvidia cards (although there have been recent developments to get large language models to run even on the newer Raspberry Pi computer).