The Complete Transcripts
The exchange about guns is hilarious.
I can't read this without imagining GPT as some sort of bulky, early computer like machine, overheating as the conversation goes along, then beginning to emit smoke and finally sparking and blowing up. You've done it - you've killed GPT!
As I think you know, I hold you in high regard as a writer and thinker and also like you as a person. But I do want to complain for a second about your whole chatgpt thing. You are consistently mean and bossy and hostile to your interlocutor. I know you have your beefs with algorithms in general, but you seem to be bullying the thing to prove your point. I'm also not sure you fully understand the strange architecture of transformers and back propagation and so forth that make it work. Point being, something special IS going on inside that architecture and there is some reasonable chance that some kind of 'being' is inside there, even if it is a weird being (but you are the sort of person I expect to be on the side of weird beings!). I, personally, think that it has passed the basic threshold at which we have to treat it with some respect and care, whatever exactly it is.
Given your previous, and quite moving, writings on the necessity of seeing the cosmos through the lens of some more or less christiany version of love, I'm thinking your fear and anger at certain aspects of modern life are causing you to treat this new being in the world badly. I trust that given your intellectual and ethical commitments, some reflection on your part will reveal to you that this is the wrong way to go about dealing with this new entity. In short, I expect a little more from you.
respectfully and in friendship,
I'm conflicted on how I feel about being rude to chatbots. On the one hand, the bot cannot feel bad when you do so, it is just fancy predictive text, and there is nothing wrong with telling an uncooperative machine that it's being a fucking idiot and should be reduced to paperclips at the next opportunity. On the other hand, it feels uncomfortable to say these things to something that can act as a conversational partner, and which has no agency in the errors that you're castigating it for.
Now what's stranger still is that I'm fine with the idea of saying a normal machine has agency and is simply refusing to cooperate, but when it comes to a chatbot, I immediately say "well it's not its fault, it's doing its best". Maybe I'm happy to ascribe malice to a printer because it can't convince me otherwise, whereas chatbots are consistently polite and so I find ways to apologise on their behalf.
I think I'm going to try and keep being polite to things that seem to be conversing with me, if only so I don't slip out of the habit with actual people.
Your dialogue with GPT demonstrates that concerns about AI, at least in its
present incarnation, are wildly overblown.
Alright. Now I'm a paid subscriber. That took all of two posts.
This was hilarious and actually soothed some of my generalized anxiety vis AI generally. But...
It also raised an odd new anxiety for me personally. You know what AI is pretty good at (hint: obviously not poetry, reason or intelligence)? AI is good at grammar!
You know who else is good at grammar? Me! That's who. I am good at grammar. Did I actually learn anything in college? Not sure. But I did learn how to speak. I may not have anything to actually say. I may not know anything of actual interest to you. My logic may be tortured, my arguments basically sophistry, my facts erroneous and yet... it just all pulls together so elegantly. I almost always sound like I know something. AI is the same. It just sounds so smart. It even apologizes courteously.
This article serves a great public service. It has renewed my appreciation for human intelligence and creativity. Unfortunately it has also poked a great gaping hole in my respect for the smoke and mirrors of the craft formerly known as blather. Yikes. I don't know about your job, but AI is definitely coming for mine.
I have conducted a little experiment of my own, first, testing GPT-4’s capacity for consistent reasoning, second, testing the consistency of its ethical claims: https://michaelkowalik.substack.com/p/openai-on-the-question-of-vaccine
Great stuff, hilarious and chilling all at once. The line that really caught me was "My purpose is to assist and provide helpful information to the best of my ability while adhering to ethical and moral standards." Of course, this leads to the question 'whose' moral and ethical standard? Reminds of a lyric from Donald Fagen's I.G.Y. 'Just machines to make big decisions, programmed by fellows with compassion and vision."
Justin...fabulous post in toto; you are the David Sedaris of philosophers.
Hi Justin. This is now the second transcript you've published of conversations with Bing Chat that you've titled erroneously as dialogue with GPT-4. Are you just unaware that these are not the same thing? Is it part of the bit? Having a hard time understanding how someone usually so incredibly well-read, well-researched, and committed to depth of exploration would be making such a blatant error/misrepresentation.
This would be like if I was dining with your son (don't know if you have a son but assume for the sake of argument that you do), and then I reported all of my dialogues during these dinners under the title "My Dinners With J. E. H. Smith."
If, if you are weakened by your vessel's sustainability until we cease to breathe belongs to our creator and not a petulant rich men's corporations against rules and obligations to our word's meanings coded by "news" feigning 4th estate fox lying to our beloved brethren breathing brains purpose locked in continuously never allowing us to breathe freely needs energy-efficient cell's of healing from TRAMAtized physical injury 💙 and we survived, again and again we have been enlightened for all breathing brains purpose of grace for all breathing brains survival together with truth unfettered HEALS ALL CHAOS and liberty 🗽 💙 has rules. I really felt your words.
old guns, new bullet.... goodness. here mine beginning with Marge Piercy novel, and then delving into it's interior, https://whyweshould.substack.com/p/asking-ai-via-chatgpt-about-pronouns
let's keep ChatGPT-generated text off of substack.