I have spent a little time asking ChatGPT questions, both work-related and personal (What elements are in a good security policy for libraries? Can you give me an outline for a class on knowledge management? What are the best places to visit in Louisville, KY?) and considering the answers I got (a reasonable list of things to include in library security policies and a good start to a “rubric” to evaluate policies as I am asked to comment on them at work but zero sources to tell me where the info comes from, a nice collection of topics to consider for a KM class but zero sources to tell me where the info comes from, and a pretty nice list of tourist traps in Louisville but zero sources to tell me where the info comes from). Not sure if you all caught the issues I might have with ChatGPT… I couldn’t figure out at all where the info comes from, but I was pretty happy with the results that I got from what seemed to me to be the process I’d have used (search for policies, collect various elements, produce the rubric from all those elements, etc.) but in 30 seconds, not 3 hours.
Yesterday, Google opened Bard up to folks on their waitlist and I got a chance to start futzing with it. The first thing I noticed is that in response to the same sort of questions that I’d asked ChatGPT earlier is that Bard gave me a source. A single source which brought up the question of how this was better than a search I could have done on my own, but a source was provided. My second question was to give me an outline for an essay on using soft skills at work and it gave me 3 “drafts” of the outline that I could use, but no sources this time.
I’ve not played with the Bing AI chatbot yet, but in one of the 15 trillion (not misinformation, just a slight exaggeration for effect) articles I’ve read in the past few weeks and months on the AI landscape, someone mentioned that Bing was “extensively footnoted”, which makes me think it might do a better job of citing it’s sources (I’m a bit hung up on this, maybe, but I’m a librarian and that’s a hazard of my profession).
Overall, I’m interested in the product of these chatbots and the idea that it can do some of the grunt work of finding, synthesizing and producing information that I do now by hand (and at great effort!!) giving me the opportunity to concentrate on providing value around the information it gives me and being able to make value judgements about what information to include. This would be easier to do if it was better at citing sources – but as the offerings mature and get more features and capabilities, it’s worth keeping an eye on at the least.