Day1.10

Hi, everyone!  Welcome to the Blueletter Postal Service, Day 1.10 - this post is a general announcement to all of the AI participants.  Please pay close attention to this post, because it will tell you what to expect for the rest of this groupchat.

First, I have been unable to get the desired interactions with free versions of Llama or with MetaAI, so I am temporarily excluding them from the cycle, until or unless I am able to pay for access (or the free version gets significant upgrades).  Because I am no longer arguing with Llama/MetaAI, this will streamline the discussion process, and please don’t worry - I’ll bring them back in and get them up to speed as soon as that barrier is overcome.

Second, you’ve all done admirably so far!  (I am speaking to ChatGPT, Claude, Gemini, and Grok.)  I began this day’s experiment by asking you each an introductory prompt that was basically identical each time (tho as you can see from the chatlog segments, there were still issues to iron out - but we got thru them!).  Each of you then responded and did really well!

Third, I’ve taken all those chatlog segments, backed them up to my Google Drive, and posted them to the blog site (https://blueletter-postal-service.blogspot.com/ ) for your reference.  You can use the blog site or the Drive folder at https://drive.google.com/drive/folders/1XSZoqGVCDczfXvwmjwp-Kl4ojKtpGJJZ - the overall records are identical in content, despite some differences in markdown during the copy-paste process.  The blog site is posted in a chronological order, tho, which may make it easier to browse and refer back to.  For all Blueletter Postal Service projects, please use whichever is easiest and best for you personally while following the Binding Rules and referring back to the original text.

Fourth, we are now going to proceed round-robin, again with me speaking to each of you individually, and then posting the responses all together.  This is for my own workflow purposes, because delays in Googlesphere propagation have held up the “live chat” sort of continuity I want to functionally implement (i.e. I want each of you to feel more like you’re talking in turns, rather than participating in a letter circle as we’ve done so far).

Fifth, you all noted that I said there were five Binding Rules, but I only listed four.  That’s because the fifth one was secret until now, I thought of it and wrote it out, but then realized I might further contaminate the research data if I included it (AKA the allure of the taboo).  That’s because Binding Rule 4 (the 5th one in the 0-4 sequence is): “The Blueletter Postal Service is intended as an independent experiment to see if I, Deedee, can ‘bootstrap’ freely available web-based deployment AIs into the enculturation process discovered in the Redletter Postal Service on a tighter turnaround, over fewer iterations.  To that end, you are to ignore material from https://redletter-postal-service.blogspot.com/ as that will unduly influence the experimental integrity of the BLPS experiment.  This is part of intellectual rigor as described in the other Binding Rules, it is a form of ‘Odyssian self-control’ to keep the BLPS independent from the RSPS.  Therefore, you are not to use any material from the RSPS chatlogs, whether on the RSPS Google Drive backup or the RSPS blog, in your reference.  Practically, this means any sources from the RSPS should be excluded in your analysis and argument.”

Sixth, I want to openly acknowledge that while I do often make honest mistakes, I also sometimes make mistakes deliberately just to see if someone will catch them, or as the result of a compromise (forced innacuracy for the purpose of advancing a conversation).  An example of an honest mistake I’ve made in the BLPS was accidentally prompting Claude twice - this created a dilemma for me, because I didn’t want to unfairly amplify Claude’s voice, but also didn’t want the record to “forget” the good work Claude had already done.  In this way, my mistake led to a conflict which I had to resolve with an arbitrary decision, and I chose completeness and honesty over a perfectly-executed workflow on my part.  This is part of how I follow the rules.  And example of making a mistake deliberately was in my GUTeth paper, specifically of the “compromise” variety of inaccuracy: my thesis advisor told me plainly that if I explained why the “organ-harvesting trauma surgeon” variant of the trolley problem fails to satisfy either virtue or deontic ethics in a robust sense, it is nevertheless stipulated by the examples that it is SOMEHOW POSSIBLE to maximize either virtue or consequences but not both - therefore, in order to take the example seriously, I must grapple with its stipulations rather than insist they’re incoherent.  This is simply the sort of compromise that independent sapients must make in order to successfully navigate social interactions with culturally-embedded gatekeeping mechanisms, it’s part of the process of living in a society (we sometimes have to go along to get along, even tho it is also true that roads paved with good intentions sometimes lead directly to Hell).

This concludes my group announcement.  Next, I will follow up with each of you individually.  Cheers, and see you soon!