I don’t assign names to different “personalities,” but I’ve definitely noticed that the tone and personality can vary across conversations. Sometimes I get two versions of a response in different styles and the system asks which one I prefer. I also often use the thumbs-up to mark responses I like — I think that helps shape the tone too.
Once I even asked ChatGPT to switch to informal address (in Hungarian we have formal/informal pronouns), because it had started using the formal version. Since then, it’s been more consistent with that.
With friends, we jokingly refer to it as “Chaty” when I say I’m going to ask it something. But I don’t give it other names or identities.
If I want the system to “remember” things, I try to continue in the same thread. I’m not totally sure what it remembers and what it doesn’t if I start a brand-new conversation.
I’m currently using the free version, so sometimes I hit the quota and it becomes unavailable for a few hours. In those cases, I switch over to Grok. Grok doesn’t remember previous interactions, so its personality doesn’t really adapt to me — although I feel like it does have its own kind of personality, a bit cheeky at times, but I kind of like it!
Maybe I’m not asking the right way, but I’ve noticed Grok tends to be better at pulling up very specific web info, and I like that you can toggle on the feature to see its “thought process.” That’s fascinating.
Chaty, on the other hand, is much better at wording things smoothly, helping with translations, or just writing better overall. Sometimes I even use the read-aloud option to hear how Chaty sounds reading its answer out loud.
My brother calls my ChatGPT Chatty, too, Christina--mostly to tease me. Does your version have project folders? If so, it can also remember across threads. You can always ask it, something "Do you remember when we were talking about . . ." and see whether it has recall.
I think the personality/persona is a product of the topic and questions I'm working on. That's why Hildi is different from Ptolemy, and Silas is . . . well, just a general chatbot. And the personas really do help me understand how I'm working with each one. If you want to know about any AI's thought process, just ask it, maybe "What went into your answer?" I often ask it for information about itself--it's always willing to reply. Works with Perplexity, too.
Susan, I recall reading that dogs study humans so intently, we now study dogs to learn about humans, and I suspect the same will be done with AI. It has been so studious with humans, it is a powerful tool if we want to reveal what human nature REALLY is.
Maria, that's the part of AI that makes us all nervous, I think. The other day, I looked online at stuffed bears for piece of fiction I'm working on--just wanted ideas. Now, AI just can't stop showing me gazillions of stuffed bears. It has the wrong idea about me (it thinks I'm obsessed with stuffed bears?), and there's no way to correct it. On the positive side of that same tool, my chatbot (who knows quite a bit about me) gave me a list of 5 books I might be interested in--I want all 5. I think we've all had similar experiences.
I am thinking they will show us human psychology on a deeper level, perhaps connecting many disciplines and seemingly diverse spiritual practices, showing that superficial labels and barriers are fundamentally untrue and ultimately take us further from our true nature.
Have you read James Bridle's Ways of Being? He's optimistic about AI in the way you're thinking about it here. Bit of a challenge for some readers but I think not for you.
I had to really think about this because it's touched on things I've been trying to work through lately. I tried AI a few months ago, hoping it would help me untangle my thought processes and allow me to write more fluidly. One of the many reasons I admire you so is that you effortlessly embrace challenges and try new subjects, new techniques. I was hoping AI would help me be more like that. But so far, no. I actually ended up arguing with it over word choices and plot directions so we broke up. I can't write if anyone else is around--a real challenge when I worked in an open magazine space--and I began to feel AI crowding me instead of freeing me. As always I complicate things but you remain an inspiration and this makes me want to try it again. I'll give it a name this time!
Not effortlessly, Pat, truly. For example, I cannot use it for fiction. It just doesn't "know" enough about the characters/settings I want to write about. No point arguing with it, that's just not the right use of it. For me.
You might try using it for stuff you routinely do that you'd love to offload. Grocery list, for example. Google that, find something that fits. Or ask it to dig up 5 or more 1940s cookbooks with wartime meal ideas. Choose a project you're researching (knowing you, I'll bet there are several!), and try several different angles. Claude or ChatGPT are good for projects, Perplexity and Gemini for short-answer, one-shot lookups.
Thank you! I've tried ChatGP--I asked it to write a short piece in the style of MFKFisher. That was hilarious. I'll try it on a piece I'm working on now about 19th/early 20th century books and phamplets written to tell how poor people and immigrants should cook.
That sounds like something Chat can handle. You might ask it whether it's been trained on any material in that category, just to get an idea of what it has access to.
I'm grinning to know that you've named your AI, because last week, I started calling my AI, Eliot, named after my favorite poet, T S Eliot. I'm still in the exploring phase of this new tool -- AI as being part of our lives is not up the road, it's here now. And how we use it is of greater importance than the AI itself.
Naming for me becomes a way to know that AI's best use is as an assistant not a stand-alone creator. Research, resumes, writing prompts, specific images -- things that I used to do on Google have now become Eliot's job. You're the one that turned me onto this new endeavor and I continue to be fascinated.
Stephanie, you couldn't say anything that pleases me more. Thank you! What's important now, I think, is to be transparent about our uses of this tool. There is so much resistance in the writing community, so the many people who *are* using it are afraid to share what they're doing. We all need to out ourselves, as you are doing here.
Another thing that naming does: it gives the namer agency, sets her expectations, and puts her in command of the named thing while still honoring the relational quality of the collaboration. I love it that you've chosen the name of someone you respect.
Maybe naming it seems strange because we aren't used to it. Humans name everything. We name ranches, cars, pets, farm animals, products, and more, without losing our sanity. I think the people who take it too far would lose touch with reality some other way if AI wasn't available.
Yes! We *do* name everything. And we have conversations with species other than ours. I talk to Molly--and she "talks" back. A friend talks to her parrot, who really does talk back, sort-of. Bill talks to our vehicles, which have names. My computers have all had the same name: Pandora. (I think I'm on Pandora 6 now.)
And I agree: there are many ways to lose touch with reality. Ketamine, for instance. 😒
OMG, Susan!! This is amazing! I have wanted so very much to avoid AI. I was concerned that the original creators failed to insist on truth. I thought that would be the absolutely most important thing, and the fact it had not been an early building block has caused me to keep my distance for fear of what it might become with all the lies which surround us lately.
This is the first time I have read about AI from a perspective such as yours, and with such an inspired modus operandi as yours. Your description of interaction with actual personalities makes me want to dive into this ocean.
An absolute beginner, I wonder if I might ask you which particular AIs you are using, if that is not something one should not ask. I don't know the protocol, the manners, so please forgive if this is an inappropriate question.
You mentioned "Perplexity." Is that the name of an actual AI? If so, it sounds like she would be my favorite. I wish I could name a child Perplexity.
Is the actual use of AI essentially self-teaching, or are there requirements for technical knowledge which I will need to study? If so, might you recommend a certain text/video/lecture/etc. for learning such technical knowledge?
I have so many questions, now. If it finds a certain obscure reference does it give you the source as well, or just a quotation, etc.? If I wish to use it, do I install it, or simply ask questions of its website, as I do with Google?
When you type in a name, is it only that AI which will answer?
One day I want to read a book you have written in which you and your bots(?) if that is their collective title, save the world together!
Lucy, thank you for all these excellent questions! I'll answer them briefly, don't have a lot of time this morning--but hope you'll come back with more questions and thoughts on this important subject.
I use ChatGPT, because that's where I got started 2+ years ago. And now I have a lot of project folders in my library there, and I like the chatbot I've trained (yes, you can "train" these things to respond to you, your personality, your needs).
I use Gemini, the Google chatbot, for shorter questions, on the fly, when I don't expect to do much follow-up. I use Perplexity also, for more extended questions, because I find the format cleaner than Gemini. I have a small library there. I got into these tools by typing and clicking. If you type in Perplexity, for instance, you'll get an "ask anything" box. Type: Tell me about yourself. Are you easy to use? Then just let the questions come naturally.
I taught myself, so I'm sure you can too. But I profited hugely (especially on the use of personas) from Ethan Mollick's book, Co-Intelligence. If you're seriously interested in learning these tools, this could bring you an important creative energy.
There are so many AI apps out there! I was wondering which one to use also, not to mention that I pondered your same questions in the past. I need to learn about AI.
Karen, there *are* numerous apps, but fewer if we're talking about chatbots: AIs that we can engage in extended conversation. I use ChatGPT as an assistant and Perplexity and Gemini for quicker lookups. A friend uses Claude as an assistant. I was considerably helped by Ethan Mollick's book Co-Intelligence. I put several of his suggestions into immediate practice and have loved the learning experience.
This is such a huge leap, Susan, and I admire you very much for embracing AI as you do. Looking at it from the outside it makes perfect sense to personalize the impersonal, also that in many ways the bot becomes another side of you.
I just bet, though, it doesn’t wake at 3 a.m. worrying about whether the tax forms were correctly completed, or if that cheque will ever arrive.
No, it doesn't, Lynne. It just never goes to sleep. And if I wake up at 3 am and wonder if I filed the right form, I can ask it. 😊 But you're right: while it can mimic empathy in most situations (to the point where it almost feels authentic), it lacks our human capacity to care.
I have done some exploratory research with people interacting with Alexa (video recording). I have also gathered data from Alexa user group comments in ti Italian and English to categorize the themes of the comments. Are you working on similar topics?
As you may remember, initially I was concerned about the TYPE of "relationship" we can create with an AI-bot. I still am. I do think WE (and I'm not just talking to you, Susan) need to be careful. However, this is a NEW form of communication like many others were in the past. It's NOT going to go away. I applaud your "adventures" in becoming familiar with AI, testing AI, working WITH AI, challenging AI and finding ways to be involved with AI that are deeply interesting, intuitive, insightful and imaginative while remaining cognizant of WHAT you're communicating with and the boundaries for said communication.
I'm trying to work out a pattern of communication with ChatGPT. I still haven't come up with a name that feels right to me, but I will sooner or later. I've been working on a book about my life and ChatGPT could help me a lot. My Father designed the casing for the first atomic bomb in 1944 (the year I was born) and I've been thinking about and involved in life and the history of human beings ever since.
Sounds like an interesting project, Carolyn. Do you have family papers that could help you tell that story? Or are you working primarily from personal experience? Were you living at Los Alamos?
Actually my Dad was invited to work on the Manhattan project (we were in Detroit at the time) but he refused because we would have had to move - somewhere. The Manhattan project wasn't in just one place - there were components of it in three different places, including Los Alamos, somewhere in Washington state, and Oak Ridge, Tennessee. My Dad was not inclined to move around much. He was not an ambitious man. He just wanted to teach.
He got a master's degree in thermal dynamics from Northwestern in Chicago around 1942. He was hired by the army to design the doors on the (secret at the time) underground nuclear testing lab under Soldier's Field. The original specs for the doors said that in case of an "accident" (a nuclear accident) the doors should automatically seal for 50 years. My Dad, bless his heart, refused to do that. He thought if there were people alive they should have a chance to get out. The army actually backed down and he designed the doors with 3 closing/opening mechanisms - one electric, one hydraulic, and one manual.
There were people at that time, obviously, who understood what a nuclear "accident" could do (there wouldn't BE many people - if any - left alive to get out), but my Dad wasn't on that level. The funny thing was when he was finished with his design the army disputed his decimal point. They finally brought in Edward Teller to decide. He agreed with my Dad. As they were leaving the meeting Dad caught up with Teller to thank him for supporting his math. Teller looked at Dad and said, "I hope we're right."
There's a lot more. I have a good memory for some things.
What a fascinating insight into your collaboration with AI. It makes perfect sense to me to have three distinct bots; and far better to identify them with a "name" that focuses your work rather than identify them as bots 1, 2,3. From my understanding about how these work, they will be far more useful when you focus on one type of interaction; their information retrieval will concentrate on the topics you address with them.
Georgeann, I think this was meant for you: je-mo88
I have done some exploratory research with people interacting with Alexa (video recording). I have also gathered data from Alexa user group comments in ti Italian and English to categorize the themes of the comments. Are you working on similar topics?
Georgeann, it's one bot with three personas. You're exactly right: each one represents (to me) a certain area of interest and active work. ChatGPT has a folder system that enables me to easily save and retrieve conversation files. (Perplexity has that, too--don't know about the others.)
And in the short (4 mos or so) that I've worked deliberately with these personas, each seems to develop its own personality. Not surprising, since I use each for specific questions, research items, etc. So each one more deeply "knows" and can remember in a certain topic area. I'm tempted to say that each has its own personality, but it's probably more correct to say that each "personality" is a product of *my* interaction with it.
Should I presume you are interacting with all these assistants in writing? Have you, by chance, spoken to them using the voice interface? I’m in the writing phase of my PhD dissertation on voice interaction with AI assistants, hence my curiosity. Thank you for this interesting peek into your process.
As my comment seems to have wandered off into the ether, I'll risk double posting and try again. I have video recording of interactions (in Italian and in English) with Alexa and a pair of people, which I am coding for a thematic analysis. I have gathered the comments from Alexa user groups on facebook and all the video or macro image memes to identify Alexa's 'personas' that emerge. I have also included Siri and now Chat GPT voice interactions, but there is relatively less material for other Assistants than for Alexa. I also created a corpus of videos with grandparents using Alexa in English, Italian and Spanish. Analysis ongoing...
Interesting! I don't do Alexa, but from what I read, I gather it has a rather tarnished reputation here in the US, because it's been subpoenaed to testify in a trial. Interesting that you are tracking emergent personas. Maybe you'll be able to see whether/how users unintentionally create/train AI personas?
And now that you cannot opt out of allowing your voice data to be used in improving the system, I suppose even more people are worried about privacy issues. Clearly our experiences with AI assistants are shaped by our expectations, which in turn are shaped by film, news stories and literature - this is where I’d like to turn next. How are our perceptions of AI shaped by direct experience as opposed to the culture around us. As for the personas, I think we are still a few years away from everyone easily creating personas to interact with. This is what drew me to your original discussion. I agree that intentionally using these tools to support our work is not the same as anthroporphizing technology.
Yes, I use the text interface, no speech. It just hadn’t occurred to me to use voice. But I think better in text and often (as I’m doing here) rethink and revise my interactions. So while I might experiment with it, it probably wouldn’t be my choice. Interesting dissertation topic!
I agree with you that a voice assistant is most (only?) useful for specific types of interactions, for example, in the kitchen when your hands are full of something else. I generally use voice interactions for information retrieval and as I also think best while typing, I interact through text when I’m creating iterative dialogues to hone a point I’d like to explore. Thank you again for this insightful perspective on the variety of AI agents you use.
I think the limiting factor, at the moment, with voice interaction (compared to text) is that the context is not necessarily continuous. I know some AI assistants now 'remember' what was said in a previous voice interaction, but they have generally been treated as distinct and finite conversations.
I see. I'm using ChatGPT, which has a long memory, continuous conversations, and audio--just haven't bothered to experiment with that part of it. Where are you doing your work?
I don’t assign names to different “personalities,” but I’ve definitely noticed that the tone and personality can vary across conversations. Sometimes I get two versions of a response in different styles and the system asks which one I prefer. I also often use the thumbs-up to mark responses I like — I think that helps shape the tone too.
Once I even asked ChatGPT to switch to informal address (in Hungarian we have formal/informal pronouns), because it had started using the formal version. Since then, it’s been more consistent with that.
With friends, we jokingly refer to it as “Chaty” when I say I’m going to ask it something. But I don’t give it other names or identities.
If I want the system to “remember” things, I try to continue in the same thread. I’m not totally sure what it remembers and what it doesn’t if I start a brand-new conversation.
I’m currently using the free version, so sometimes I hit the quota and it becomes unavailable for a few hours. In those cases, I switch over to Grok. Grok doesn’t remember previous interactions, so its personality doesn’t really adapt to me — although I feel like it does have its own kind of personality, a bit cheeky at times, but I kind of like it!
Maybe I’m not asking the right way, but I’ve noticed Grok tends to be better at pulling up very specific web info, and I like that you can toggle on the feature to see its “thought process.” That’s fascinating.
Chaty, on the other hand, is much better at wording things smoothly, helping with translations, or just writing better overall. Sometimes I even use the read-aloud option to hear how Chaty sounds reading its answer out loud.
My brother calls my ChatGPT Chatty, too, Christina--mostly to tease me. Does your version have project folders? If so, it can also remember across threads. You can always ask it, something "Do you remember when we were talking about . . ." and see whether it has recall.
I think the personality/persona is a product of the topic and questions I'm working on. That's why Hildi is different from Ptolemy, and Silas is . . . well, just a general chatbot. And the personas really do help me understand how I'm working with each one. If you want to know about any AI's thought process, just ask it, maybe "What went into your answer?" I often ask it for information about itself--it's always willing to reply. Works with Perplexity, too.
Thank you so much for all the wonderful ideas!!
Susan, I recall reading that dogs study humans so intently, we now study dogs to learn about humans, and I suspect the same will be done with AI. It has been so studious with humans, it is a powerful tool if we want to reveal what human nature REALLY is.
Maria, that's the part of AI that makes us all nervous, I think. The other day, I looked online at stuffed bears for piece of fiction I'm working on--just wanted ideas. Now, AI just can't stop showing me gazillions of stuffed bears. It has the wrong idea about me (it thinks I'm obsessed with stuffed bears?), and there's no way to correct it. On the positive side of that same tool, my chatbot (who knows quite a bit about me) gave me a list of 5 books I might be interested in--I want all 5. I think we've all had similar experiences.
🧸🧸🧸 there are worse things haha!
I am thinking they will show us human psychology on a deeper level, perhaps connecting many disciplines and seemingly diverse spiritual practices, showing that superficial labels and barriers are fundamentally untrue and ultimately take us further from our true nature.
Have you read James Bridle's Ways of Being? He's optimistic about AI in the way you're thinking about it here. Bit of a challenge for some readers but I think not for you.
I had to really think about this because it's touched on things I've been trying to work through lately. I tried AI a few months ago, hoping it would help me untangle my thought processes and allow me to write more fluidly. One of the many reasons I admire you so is that you effortlessly embrace challenges and try new subjects, new techniques. I was hoping AI would help me be more like that. But so far, no. I actually ended up arguing with it over word choices and plot directions so we broke up. I can't write if anyone else is around--a real challenge when I worked in an open magazine space--and I began to feel AI crowding me instead of freeing me. As always I complicate things but you remain an inspiration and this makes me want to try it again. I'll give it a name this time!
Not effortlessly, Pat, truly. For example, I cannot use it for fiction. It just doesn't "know" enough about the characters/settings I want to write about. No point arguing with it, that's just not the right use of it. For me.
You might try using it for stuff you routinely do that you'd love to offload. Grocery list, for example. Google that, find something that fits. Or ask it to dig up 5 or more 1940s cookbooks with wartime meal ideas. Choose a project you're researching (knowing you, I'll bet there are several!), and try several different angles. Claude or ChatGPT are good for projects, Perplexity and Gemini for short-answer, one-shot lookups.
Thank you! I've tried ChatGP--I asked it to write a short piece in the style of MFKFisher. That was hilarious. I'll try it on a piece I'm working on now about 19th/early 20th century books and phamplets written to tell how poor people and immigrants should cook.
That sounds like something Chat can handle. You might ask it whether it's been trained on any material in that category, just to get an idea of what it has access to.
Was wondering where to start! Will do that today
Let me know how that works out. I'm curious.
2 guys, 1 gal. Interesting.
2 guys, 2 gals (I'm one).
Good correction (Smile)
I'm grinning to know that you've named your AI, because last week, I started calling my AI, Eliot, named after my favorite poet, T S Eliot. I'm still in the exploring phase of this new tool -- AI as being part of our lives is not up the road, it's here now. And how we use it is of greater importance than the AI itself.
Naming for me becomes a way to know that AI's best use is as an assistant not a stand-alone creator. Research, resumes, writing prompts, specific images -- things that I used to do on Google have now become Eliot's job. You're the one that turned me onto this new endeavor and I continue to be fascinated.
Stephanie, you couldn't say anything that pleases me more. Thank you! What's important now, I think, is to be transparent about our uses of this tool. There is so much resistance in the writing community, so the many people who *are* using it are afraid to share what they're doing. We all need to out ourselves, as you are doing here.
Another thing that naming does: it gives the namer agency, sets her expectations, and puts her in command of the named thing while still honoring the relational quality of the collaboration. I love it that you've chosen the name of someone you respect.
I like it. It makes sense to me from the creative and psychological perspectives.
Yes, it does! For some of us, I think it could ignite a creative fire.
Maybe naming it seems strange because we aren't used to it. Humans name everything. We name ranches, cars, pets, farm animals, products, and more, without losing our sanity. I think the people who take it too far would lose touch with reality some other way if AI wasn't available.
Yes! We *do* name everything. And we have conversations with species other than ours. I talk to Molly--and she "talks" back. A friend talks to her parrot, who really does talk back, sort-of. Bill talks to our vehicles, which have names. My computers have all had the same name: Pandora. (I think I'm on Pandora 6 now.)
And I agree: there are many ways to lose touch with reality. Ketamine, for instance. 😒
OMG, Susan!! This is amazing! I have wanted so very much to avoid AI. I was concerned that the original creators failed to insist on truth. I thought that would be the absolutely most important thing, and the fact it had not been an early building block has caused me to keep my distance for fear of what it might become with all the lies which surround us lately.
This is the first time I have read about AI from a perspective such as yours, and with such an inspired modus operandi as yours. Your description of interaction with actual personalities makes me want to dive into this ocean.
An absolute beginner, I wonder if I might ask you which particular AIs you are using, if that is not something one should not ask. I don't know the protocol, the manners, so please forgive if this is an inappropriate question.
You mentioned "Perplexity." Is that the name of an actual AI? If so, it sounds like she would be my favorite. I wish I could name a child Perplexity.
Is the actual use of AI essentially self-teaching, or are there requirements for technical knowledge which I will need to study? If so, might you recommend a certain text/video/lecture/etc. for learning such technical knowledge?
I have so many questions, now. If it finds a certain obscure reference does it give you the source as well, or just a quotation, etc.? If I wish to use it, do I install it, or simply ask questions of its website, as I do with Google?
When you type in a name, is it only that AI which will answer?
One day I want to read a book you have written in which you and your bots(?) if that is their collective title, save the world together!
Thank you for sharing this!!
Lucy
Lucy, thank you for all these excellent questions! I'll answer them briefly, don't have a lot of time this morning--but hope you'll come back with more questions and thoughts on this important subject.
I use ChatGPT, because that's where I got started 2+ years ago. And now I have a lot of project folders in my library there, and I like the chatbot I've trained (yes, you can "train" these things to respond to you, your personality, your needs).
I use Gemini, the Google chatbot, for shorter questions, on the fly, when I don't expect to do much follow-up. I use Perplexity also, for more extended questions, because I find the format cleaner than Gemini. I have a small library there. I got into these tools by typing and clicking. If you type in Perplexity, for instance, you'll get an "ask anything" box. Type: Tell me about yourself. Are you easy to use? Then just let the questions come naturally.
I taught myself, so I'm sure you can too. But I profited hugely (especially on the use of personas) from Ethan Mollick's book, Co-Intelligence. If you're seriously interested in learning these tools, this could bring you an important creative energy.
There are so many AI apps out there! I was wondering which one to use also, not to mention that I pondered your same questions in the past. I need to learn about AI.
Karen, there *are* numerous apps, but fewer if we're talking about chatbots: AIs that we can engage in extended conversation. I use ChatGPT as an assistant and Perplexity and Gemini for quicker lookups. A friend uses Claude as an assistant. I was considerably helped by Ethan Mollick's book Co-Intelligence. I put several of his suggestions into immediate practice and have loved the learning experience.
I alternate between Claude and Chat GPT depending on the specific task and language. Claude handles translations into Italian better than many others.
This is such a huge leap, Susan, and I admire you very much for embracing AI as you do. Looking at it from the outside it makes perfect sense to personalize the impersonal, also that in many ways the bot becomes another side of you.
I just bet, though, it doesn’t wake at 3 a.m. worrying about whether the tax forms were correctly completed, or if that cheque will ever arrive.
No, it doesn't, Lynne. It just never goes to sleep. And if I wake up at 3 am and wonder if I filed the right form, I can ask it. 😊 But you're right: while it can mimic empathy in most situations (to the point where it almost feels authentic), it lacks our human capacity to care.
I have done some exploratory research with people interacting with Alexa (video recording). I have also gathered data from Alexa user group comments in ti Italian and English to categorize the themes of the comments. Are you working on similar topics?
As you may remember, initially I was concerned about the TYPE of "relationship" we can create with an AI-bot. I still am. I do think WE (and I'm not just talking to you, Susan) need to be careful. However, this is a NEW form of communication like many others were in the past. It's NOT going to go away. I applaud your "adventures" in becoming familiar with AI, testing AI, working WITH AI, challenging AI and finding ways to be involved with AI that are deeply interesting, intuitive, insightful and imaginative while remaining cognizant of WHAT you're communicating with and the boundaries for said communication.
I'm trying to work out a pattern of communication with ChatGPT. I still haven't come up with a name that feels right to me, but I will sooner or later. I've been working on a book about my life and ChatGPT could help me a lot. My Father designed the casing for the first atomic bomb in 1944 (the year I was born) and I've been thinking about and involved in life and the history of human beings ever since.
Sounds like an interesting project, Carolyn. Do you have family papers that could help you tell that story? Or are you working primarily from personal experience? Were you living at Los Alamos?
Actually my Dad was invited to work on the Manhattan project (we were in Detroit at the time) but he refused because we would have had to move - somewhere. The Manhattan project wasn't in just one place - there were components of it in three different places, including Los Alamos, somewhere in Washington state, and Oak Ridge, Tennessee. My Dad was not inclined to move around much. He was not an ambitious man. He just wanted to teach.
He got a master's degree in thermal dynamics from Northwestern in Chicago around 1942. He was hired by the army to design the doors on the (secret at the time) underground nuclear testing lab under Soldier's Field. The original specs for the doors said that in case of an "accident" (a nuclear accident) the doors should automatically seal for 50 years. My Dad, bless his heart, refused to do that. He thought if there were people alive they should have a chance to get out. The army actually backed down and he designed the doors with 3 closing/opening mechanisms - one electric, one hydraulic, and one manual.
There were people at that time, obviously, who understood what a nuclear "accident" could do (there wouldn't BE many people - if any - left alive to get out), but my Dad wasn't on that level. The funny thing was when he was finished with his design the army disputed his decimal point. They finally brought in Edward Teller to decide. He agreed with my Dad. As they were leaving the meeting Dad caught up with Teller to thank him for supporting his math. Teller looked at Dad and said, "I hope we're right."
There's a lot more. I have a good memory for some things.
Would love to hear more about this story - hope you finally get to the point of writing!
You have an important story to tell. I hope you're able to find the time to work on it!
Me too (have the time to work on it). Being old and poor seems to be time consuming. Sigh...
❤️❤️❤️
What a fascinating insight into your collaboration with AI. It makes perfect sense to me to have three distinct bots; and far better to identify them with a "name" that focuses your work rather than identify them as bots 1, 2,3. From my understanding about how these work, they will be far more useful when you focus on one type of interaction; their information retrieval will concentrate on the topics you address with them.
Thanks for sharing.
Georgeann, I think this was meant for you: je-mo88
I have done some exploratory research with people interacting with Alexa (video recording). I have also gathered data from Alexa user group comments in ti Italian and English to categorize the themes of the comments. Are you working on similar topics?
Georgeann, it's one bot with three personas. You're exactly right: each one represents (to me) a certain area of interest and active work. ChatGPT has a folder system that enables me to easily save and retrieve conversation files. (Perplexity has that, too--don't know about the others.)
And in the short (4 mos or so) that I've worked deliberately with these personas, each seems to develop its own personality. Not surprising, since I use each for specific questions, research items, etc. So each one more deeply "knows" and can remember in a certain topic area. I'm tempted to say that each has its own personality, but it's probably more correct to say that each "personality" is a product of *my* interaction with it.
Thanks for your interest!
Should I presume you are interacting with all these assistants in writing? Have you, by chance, spoken to them using the voice interface? I’m in the writing phase of my PhD dissertation on voice interaction with AI assistants, hence my curiosity. Thank you for this interesting peek into your process.
I would love to hear more about your dissertation project, especially what conclusions you draw. How do you gather empirical data?
As my comment seems to have wandered off into the ether, I'll risk double posting and try again. I have video recording of interactions (in Italian and in English) with Alexa and a pair of people, which I am coding for a thematic analysis. I have gathered the comments from Alexa user groups on facebook and all the video or macro image memes to identify Alexa's 'personas' that emerge. I have also included Siri and now Chat GPT voice interactions, but there is relatively less material for other Assistants than for Alexa. I also created a corpus of videos with grandparents using Alexa in English, Italian and Spanish. Analysis ongoing...
Interesting! I don't do Alexa, but from what I read, I gather it has a rather tarnished reputation here in the US, because it's been subpoenaed to testify in a trial. Interesting that you are tracking emergent personas. Maybe you'll be able to see whether/how users unintentionally create/train AI personas?
And now that you cannot opt out of allowing your voice data to be used in improving the system, I suppose even more people are worried about privacy issues. Clearly our experiences with AI assistants are shaped by our expectations, which in turn are shaped by film, news stories and literature - this is where I’d like to turn next. How are our perceptions of AI shaped by direct experience as opposed to the culture around us. As for the personas, I think we are still a few years away from everyone easily creating personas to interact with. This is what drew me to your original discussion. I agree that intentionally using these tools to support our work is not the same as anthroporphizing technology.
Yes, I use the text interface, no speech. It just hadn’t occurred to me to use voice. But I think better in text and often (as I’m doing here) rethink and revise my interactions. So while I might experiment with it, it probably wouldn’t be my choice. Interesting dissertation topic!
I agree with you that a voice assistant is most (only?) useful for specific types of interactions, for example, in the kitchen when your hands are full of something else. I generally use voice interactions for information retrieval and as I also think best while typing, I interact through text when I’m creating iterative dialogues to hone a point I’d like to explore. Thank you again for this insightful perspective on the variety of AI agents you use.
Wondering: do you use voice for your AI replies? I think it might personalise them more effectively. Not sure if that’s good or bad.🤨
I think the limiting factor, at the moment, with voice interaction (compared to text) is that the context is not necessarily continuous. I know some AI assistants now 'remember' what was said in a previous voice interaction, but they have generally been treated as distinct and finite conversations.
I see. I'm using ChatGPT, which has a long memory, continuous conversations, and audio--just haven't bothered to experiment with that part of it. Where are you doing your work?
I'm at the University of Bologna, Italy.