4 ms·
The Wikipedia approach is exactly what I’m talking about, and its telling that Wikipedia's sources for this are… news articles.
by mbgerring 3y ago
The Wikipedia approach is exactly what I’m talking about, and its telling that Wikipedia's sources for this are… news articles.
- ProllyInfamous 3y agoI've held on to my 2002 Wikipedia account, and have been SCREAMING "this is the way" since way back when Microsoft was still trying to sell an encyclopedia [on CD-ROM!]. I remember industry "experts" in my own academic village questioning "but how will you TRUST something that anybody can edit?" ---- January 2023, I finally had an LLM cite one of my own written contributions when I began asking about a nuanced technical database which I maintain the article on wikipedia (which was really cool/eerie). This was over a year ago, and every day I continue to be floored by genAI capabilities in discussing published dataset complexities. I know several lowkey published authors (one whose autobiography I copyedited) that have been terrified by what LLMs know about their own lives, based on commonsense associations/deductions.
- dredmorbius 3y agoI've had a GPT (FastGTP from Kagi) cite one of my own HN contributions back to me, which was somewhat surreal. The prompt had not specifically mentioned either HN or my username. This makes me think that the depth of training material on at least certain topics is thinner than might be hoped.