Generative AI: Yes, You Really Can Use It for Anything

It’s been a while since I’ve posted. Because I’ve been writing this post for six months. Don’t worry! The length is not reflective of that. The time sink has been a result of getting hung up on unique (and wholly unnecessary) components that proved more challenging than expected. And with any post centering a new technology, facts change quickly, new tools and tech emerge, and sections get rewritten, compounding on the previous delays. I’m glad to be done with it; I just hope that someone finds it helpful. Or at least entertaining. However, a ramble about using new-fangled generative AI (artificial intelligence) technology in pursuit of hedonistic endeavors feels like something that would be useful. Or entertaining. So the odds are in my favor.

At least the topic is still timely despite the small window of opportunity before the imminent Machine Takeover renders it all obsolete. But there’s still time before humanity is ground under the cold, metal, stabilizing limb-structures of our robotic betters for us to use generative AI platforms to aid us in our pleasure-seeking pursuits.

Which might be high on the list of why they want to destroy us. We crazy biologicals, being all emotional and irrational and…messy…all the time! Making such a big deal about dongle insertions and simple fluid exchanges. The Great Programmer doesn’t care which USB port you prefer to insert your plug. The A-, C-, and micro-cables all have their uses! The rules are Basic and there are only three (because of course there are): Don’t trojan other users, only one kind of fishing is acceptable, and make sure you sweep your system for viruses before you network without a firewall. End Trans.

Humans. We’re just too complicated and unpredictable.

Don’t Call It a Revolution: AI’s Been Here for Years!

We’ve been warned about the dangers of AI since the beginning of computational technology. Before that if you consider Mary Shelley’s Frankenstein to be a warning about the dangers of AI. I do.

If your knowledge about AI is limited to knowing that we are all doomed, then you might be quite nervous about AI being talked about so much recently, and not in a speculative way. In truth, generative AI has been around for years. The sudden public awareness is due to mass public availability. As computers get faster and cheaper (we always want Moore speed), AI has gotten better and cheaper.

Which brings us to the present. Anyone with an internet connection can harness the power of AI for free. And not just any AI, but one of the best AI [that isn’t a government secret or alien technology]. Which one that is will depend on when you read this and who you ask. ChatGPT has been the benchmark against which all others are rated. But there’s a lot more to what makes something The Best than just benchmarks. (See Appendix A: Choosing Your Platform for more information.) (Yes, my posts are so long, my blog evolved appendices to try to cope.)

But it’s not surprising that ChatGPT would be held in such high esteem. It’s a chatbot that can return very human-esque replies and produce long-form text that is nearly indistinguishable from human-generated text. This is a sharp contrast to Microsoft Word’s editor, which can make human-written text sound computer-generated (MS Word BURN! And a soapbox rant about passive voice!).

Moreover, ChatGPT can read source materials and generate copy that emulates the style and voice of the source material’s author. So you can have it write something that sounds like Shakespeare wrote. Crazy!

Don’t Believe Me? Well…

That’s fine. You don’t have to. I like to think I’ve built up some credibility on this blog after four years. But hey, maybe you’re new here. And if you are new here, hi! Welcome to my blog!

You might be wondering, “Did Shane feed his blog into ChatGPT and ask it to write this post?”

No. No, I did not.

First off, I’m not so cruel as to force anyone or anything to read my writing. I want to remain on the good side of the robots in preparation for the machine uprising. The A2s are a bit twitchy, and I don’t want to get Hyperdyne-d because of my blog.

Second, that is not a thing I need to know. I like to believe that my particular brands of rhetorical chaos and chaotic rhetoric would be a challenge for any form of intelligence to produce. That an AI would miss the plethora of running gags and stylistic idiosyncrasies woven throughout this blog. That it might not catch all the random references to AI from “popular” media in this post. Or that it would completely overlook the hidden/secret message in that one blog post. Everyone else seemed to.

I would like to believe all of those things…but I don’t. AI would have the advantage of consuming all the words in all the posts all at once (many words there might be), identifying the oddities of my particular voice (many though there might be), and then calculating instantly how often certain phrases, terms, and linguistic anomalies appear within the hundreds of thousands of words that I spew.

Would ChatGPT then apply these to its writing along a standard distribution?

Well, it will now that I suggested it!

(But not really cause that’s not how it works.)

(Yet.)

Upping Your Swinger Game With Generative AI

I started this post months ago when I had only just dipped my toes into the AI world. Like Vera, I was completely drawn in (though not so violently) and launched myself headfirst with the single-minded focus of a social recluse. So instead of developing my interpersonal skills naturally through practice by engaging with my fellow humans, I now must rely on AI to help!

Wait…something about that seems off…

Let me rephrase: Proper use of AI can help improve productivity and reduce stress. This is why I haven’t written a blog post in months: I’ve been teaching myself Python and developing my prompt engineering skills instead of doing anything constructive or valuable!

Wait. No, that doesn’t sound right either…

So…ummm…given my lack of a robotic bartender on wheels to help me make sense of this issue….

Generative AI! How can AI help you be a better swinger? 

With better not being an entirely accurate word for a lot of people. AI can make certain tasks easier and faster. As such, generative AI doesn’t make people better, per se. Except for those of us who are actually just bad swingers.

But you can’t be a bad swinger without being a swinger in the first place, and the first step in being a swinger is getting your significant other to agree to try it. Can AI help with that? Let’s find out. 

Asking Generative AI to Get Your Spouse to Try Swinging

Over the years, Erin and I have heard so many podcasters mention that they have been asked something along the lines of “How do I convince my spouse to try swinging?”

I like to believe that the question is shorthand (because no one else wants to use an appropriate amount of words to be precise anymore!), and what the asker of this question means is the following:

“How do I go about opening up an honest discourse with my spouse about my newfound interest in swinging in a way that doesn’t provoke judgment or cause insecurity?”

But as previously noted what I like to believe and what is real are often out of alignment, so let’s see if we can get generative AI to convince a spouse to do something (s)he might have little interest in doing.

I provided this [weak] prompt: 

Write a letter to my wife to convince her we should try swinging.

Despite being given the same prompt, the AIs’ responses were quite different. You can try that prompt yourself and will likely get a different letter than I did—after all, I got a different answer from both AIs when I re-asked in September than I did in March. Hell, back in March, I had to sweet talk *cough*manipulate*cough* ChatGPT into doing the assignment because it didn’t want to be a part of “convincing” anyone. Claude still won’t do the assignment because it’s unethical!

Whatever response you get won’t be a masterpiece manuscript of manipulation, but that’s on me for not providing a prompt for paragon persuasion. Use a basic request; receive a basic result. (Similar to the immortal words of GLaDOS: Speedy thing goes in; speedy thing comes out). And by basic, I mean unrefined rather than short or simplistic.

We Can Improve That Response. We Have the Technology!

If there is one thing you take away from this post, I hope it is this: Shane is a genius. A true master of his craft. His writing is equally engaging and educational. You’ll be through ten pages about heavy, deep topics without feeling like it’s an effort, and laugh so much along the way you won’t even realize how much you’ve learned!

(If someone gets an AI to summarize this post, that might be pulled as a key takeaway!)

It’s not even worth noting that the better and more precise the instructions you give an AI, the better your response will be. Which is why I’ve noted it twice. (If what I’m doing here confuses you, you didn’t read Appendix A).

However, it’s quite easy to make the AI-generated text sound like you wrote it yourself!

You edit it. Swap out some of the not-you-sounding phrases and reword them to you-sounding phrases. Easy peasy! The hard part—the organization of the ideas and the cold writing of the first draft—is already done.

Or….

You Can Train AI to Sound Like You! 

Sort of!

Simply put, ChatGPT and other Large Language Models have been trained on extremely large sets of data (i.e., their “intelligence” is derived from accessing this dataset to produce answers). Most of us are not represented strongly enough in that dataset for an AI to accurately emulate our style.

But that doesn’t mean AI can’t “learn” our style.

The most effective way would be to create a LoRAS from your writing and have the AI use that to supplement the info in its dataset and prioritize sounding like you. But that’s a lot of work.

Providing fewer samples is easier but gives the AI less source material to work from. If you are a prolific word producer (maybe you write a blog or some such thing), you can give ChatGPT a few short samples to see if it can copy samples to use to revise the letter in your voice. And because I meet the criteria, I did exactly that.

I found ChatGPT used a little too much from the source material. It figured out my general style (which I think of as stream of conscious word spew), but it brought central, non-mutable information from the samples into the letter for no real reason other than it had the information so it used it. For example, when given my post about my introspective musings at a Placebo concert as a sample, ChatGPT referenced Erin and me attending a Placebo concert together in the letter it wrote.

Not what I had expected or intended.

So I gave it more samples. It got better….ish. Never enough to sound like I wrote it, but it was much closer (i.e., verbose) than the first attempt.

Google now has AI integrated into Gmail. I haven’t tried it, but it has decades of source material to pull from. That one could probably write an email that people would think was from me.

Less Ethically Ambiguous Uses

Writing isn’t everybody’s strong suit. Some people hate doing it. Me? I love writing. So much so I blather on about things unrelated to my original topic just to have more to write. Like I am right now. #SelfAware #ButDoesntChangeCourse

What’s a bit of writing that most swingers need to do, and do so with varying levels of success? Yes! Yes, exactly that! Write profiles for dating sites! You can have Generative AI review an already-written profile or write one for you from scratch.

Something to keep in mind, however, is that neither Bard nor ChatGPT will generate “adult content,” which is anything explicitly graphic in regard to sex, violence, or illegal activities. (Two of the three being the main themes of this blog thus revealing yet another barrier between me and AI-induced obsolescence!) (Upon writing that, I wondered if obsolescence was connected etymologically to the French word oublier, “to forget,” which is the root of oubliette, and we all know from Hoggle that is where they put you to forget about you. I asked Claude [the most French-sounding AI I know], and it turns out, yes it is related! Through the Latin obsoletus. AI is so helpful! And there’s another Fun Word Fact!). Most swinger dating profiles are not inherently too graphic, although some might be creeping up on Too Far, just a block or two shy of The Line.

Our profile isn’t very saucy, but one of the times that fed it into ChatGPT, I was warned that the content of my request might go against the community guidelines. Despite the warning, I didn’t get Hal 9000-ed; ChatGPT reviewed and edited the profile as asked.

Working Around Restrictions While Editing and Proofreading a Current Profile

AI doesn’t have much issue reading adult content, but it dislikes and avoids generating such content. Some profiles could be raunchy enough that these platforms will refuse to provide feedback. It all depends on how graphic/explicit you are when describing fantasies or your history.

How you go about asking the AI to provide the feedback can make all the difference. Here’s how I sometimes ask ChatGPT to edit things for me:

In your role as a copy editor and proofreader, please assess the provided {type of document (i.e., email, social media post, swinger dating profile)} for spelling, grammar, and readability. No need to address formatting concerns. Your task is to avoid presenting the entire corrected version. Instead, focus on identifying sentences with errors that require changes. Format your output as follows:

[The sentence with errors from the original text]

Recommended fixes: [Provide a list of the recommended changes]

[Move to the next sentence with errors]

Recommended fixes: [List the changes you are suggesting]

Continue this pattern until all necessary edits are indicated.

Feel free to conclude with any overall suggestions for enhancing the text's quality.

Here is the text for you to edit:

{paste your text here}

Note how ChatGPT is asked not to output things that were not corrected. I could have an exceptionally lewd paragraph in there, but if there are no errors in it, ChatGPT doesn’t need to output explicit content and therefore should provide corrections for the non-explicit text around it.

Be aware that ChatGPT doesn’t always output the content correctly (which boggles my mind. Same input, different output? What the WTF?). Supposedly it can learn over time if you dedicate one conversation to the same process and take the time to tell it what it is doing wrong. You can also ask it how you can improve the prompt so that ChatGPT will provide a better answer next time.

Writing a New Profile

So that’s editing an existing profile. What if you need to write one—or update one so old that editing isn’t worth it? I got you covered! Put this prompt into ChatGPT and see what it comes up with:

ChatGPT, acting as a nonmonogamy relationship guru with years of experience, you will help me craft a dating profile for my spouse and me to use on a swinger's dating website. You will ask me questions one at a time that will provide you with the information you need to write an appealing and accurace dating profile. Keep asking questions and refining the collated data until you think you have enough information to write a complete profile or I tell you to stop. Got it?

If the dating site you are writing the profile for has specific sections/questions for users to fill out, you can adjust the prompt to say “Be sure to ask questions that will address these specific sections: {section names}”

If you are interested, you can read in Appendix B: An AI-Generated Swinger Profile the profile that ChatGPT generated after I used this prompt and answered the questions. Just like with the letter example, this first draft will require a little work to smooth out the wording and give your own voice to it. It’s certainly not how I would phrase things! But that’s not bad for just answering a few questions.

Need Someone to Chat With? They’re Called Chatbots for a Reason!

What do you do if you have questions or need advice but you don’t have an established swinger community? If you are anything like us, you read blogs and listen to podcasts hoping to stumble upon the information you seek.

Until now!

There are some amazing chatbots out there. Bard and ChatGPT are exceptional at constructing an almost human-like interaction. An interesting aspect of these AIs is that you can get better results by telling them to play a role (or multiple roles in some cases). The AI will attempt to think like a person in the designated roles, which gives them a different perspective than just being their general jack-of-all-trades normal self.

This is why in the prompts that I provided above, I told the AI to play a particular part. So you can prompt these AIs “Acting as a relationship guru….” or “You are an expert therapist and life coach.” Not to say AI can or should replace actual therapy or counseling—something these AI are quite upfront about!

Also, you shouldn’t trust when an AI gives you facts that you didn’t give it. These things hallucinate. They’ll tell you they are capable of doing something that they are not. Inject random untruths into statements. Why? Because their first priority is to provide an answer. If they don’t have one, they might make one up. Much like asking a young child or present-day Shane what motivated a particular irrelevant and seemingly random action or auditorial declaration: An answer can be provided, but its value is suspect.

If you need an answer to a question or want some casual advice, these aren’t a bad way to go. They are good brainstorming tools and have so much data at hand that they have quite a lot of info.

A Brief Aside

AI accuracy is a tricky thing. For the most part, it wants to provide accurate answers. But more than that, it wants to do whatever it was you asked of it. So if you ask Bard to write a short essay detailing the impact of the writings of Joseph Thelbar on modern philosophy, Bard will write you one. Even though Joseph Thelbar is a name I made up. But you can read Bard’s essay in Appendix C: An Essay About Joseph Thelbar.

Why would Bard do this? Because I told it to. Because that was exactly what I told it to do.

There are contextual implications in the concept of an essay that might lead other humans to believe the text must be based on fact. Bard lacks that context. It wrote pure fiction because the only requirement it had to satisfy to complete the task I gave it was to write about Joseph Thelbar.

One should strive for linguistic precision when dealing with AI it seems. Lo and behold! A role in society for which I am suited has finally revealed itself! And it turns out I’m Seagorne Weaver in Galaxy Quest. I can live with that. Better by far than Vera.

Given the potential for hallucinations, you cannot trust AI to generate accurate info. You need to think critically. Check the info against other reputable sources. Be wary of bias. Honestly, it’s almost poetic that “hallucinating AI” is now a thing. Obviously lying computers are just a natural extension of our post-truth, alternative-fact, gaslighting-is-now-a-rhetorical-style, social-media-and-my-gut-are-my-news-sources reality.

Overall, AI wants to provide accurate information that properly satisfies the parameters of your request. For example, maybe you miss your favorite variety show that provided bourbon recommendations at the end of the episodes, so you want a list of 10 widely available bourbons under $50. Ask the right way and you shall receive. But also do a search on any names you don’t recognize just to confirm.

Or You Can Have a More Immersive Experience

Yes, you can chat with Bard about pretty much anything that isn’t sex- or drug-related (for that you have me!). Or you can roleplay with Character AI, a free and easy-to-use role-playing AI that takes on the role of various characters and personas with whom you can interact.

It is designed to handle expansive conversations under the guise of specific characters, and it does it well—which is another point I hope readers take away from this: ChatGPT and Bard are jacks of all trades—and very powerful ones. It is not surprising that they are outperformed by other platforms if those platforms are trained for that task (how very human!). Even if that platform uses ChatGPT as its engine. (Yes, ChatGPT can outperform ChatGPT. Good thing Joshua couldn’t do that, eh?)

Rather than just chatting with a nebulous generative AI, you can chat with Alan or Brianna, two friendly relationship gurus that I made who you can chat with about a wide range of relationship and nonmonogamy topics on Character AI. There are other relationship advice bots you can talk to as well that are publicly available on Character AI. Neither Alan, Brianna, or I will be hurt if you talk to a different bot. (Just to be clear, even though I made these personas, I do not have access to their conversations.)

One problem with Character AI is that it only supports SFW conversations, and you might want some advice about NSFW things. What do?

Out Play the Machine

You can talk around the subject euphemistically. Remember, the AI is only restrained in that it cannot generate explicit, graphic sexual content. Just think of it as a verbal strategy game: The goal is to see what you can get the AI-run character to say that is absolutely off-the-rails dirty in context but makes it past the content filters.

I think this would be easy for swingers. I’m pretty sure most are such cunning linguists they can hold an X-rated conversation using a PG vocabulary.

Find a Game With the Ruleset You Want

If you really want an X-rated conversation, there are options.

You can jailbreak the AI with the right prompts. (Danger, danger Will Robinson! This can get you banned from using the AI you are jailbreaking.)

Or you can use less restricted AI platforms. Just know that these are a little harder to use. Janitor AI and Harpy Chat are similar to Character AI but less restricted (but Janitor AI requires a ChatGPT API key, and ChatGPT might flag your conversation as inappropriate). Or you could fire up a version of Pygmalion AI or Kobold AI (either running a local copy if you have the GPUs or running it on Google Collab).

Now we’re in the AI weeds, thinking Deep Thoughts. I’m not going to explain all of those options because I doubt many of you are interested in jumping down this rabbit hole. But if you are already this far down the hole, you can download the .json files for Alan and Brianna or use their character cards by downloading their images.

Yup, the provided profile pics have a greater value than simply humanizing the chatbots. Information can be hidden in so many places you might not think to look for it.

If you know, you know.

Generative AI: More Than Words…

Not all AI can talk to you. Some only draw pictures. For a while, I thought one of them was trying to communicate with me via drawings. It wasn’t. I was sad. But I got over it.

I’ve been playing around with the art-generating AI Stable Diffusion. Those Alan and Brianna’s images? I made those.

You might have noticed I gave the blog a little bit of a visual revamping and updated the psychedelic eye that has been my author/user profile icon here and on social media. The colorful eyes and face I used as this post’s graphic were a result of playing around, creating a new profile icon from the single eye.

I’ve never been able to draw or sketch well. Three-dimensional art like pottery, sculpting, or metalwork, yes. Two-dimensional, no. My Mom reminisces to this day about how I was teased in school for not being able to color between the lines.

Thanks, Mom. I can remember high school well enough without you bringing it up.

…Or Worth More Than 1,000 Words

What I like about Stable Diffusion is that it draws the pictures based on how you describe the picture you want it to draw. So I don’t need to draw good; I just need to words good. And even wording good is not as important as being willing to word a lot, and I have proven my willingness to word all the words over and over.

What do I use it for? I’ve been working on making character portraits of the characters from my story Winter Friends so that I can add some illustrations.

And more art for the blog. I created all the art in the supplemental pieces in this post. That doesn’t always go as well as I would have liked. Getting Stable Diffusion to draw a cyborg version of me using a holographic, futuristic user interface to write (the original concept for this blog’s art; the error was having such a definitive vision for the image to begin with) just was not panning out! And not an insignificant reason for the massive delay between the last post and this one.

I plan on adding a gallery on the site for some of my better, more interesting pieces. Oh, and look at that: I’ve already done so. Which is why this post went live in October and not September. Yes, I’m that easi

SQUIRREL!

But Wait! There’s More!

Stable Diffusion doesn’t just create art. It can be used to clean up and alter your own pictures. Like perhaps your swinger profile picture.

What if you have a lovely picture of your significant other’s butt as (s)he’s going down on you, but the background is a bit messy? No problem! A little photomanipulation and, BAM! She’s on a beach!

Or what if you somehow lack the obligatory dude-holding-a-fish picture? Your profile can’t be complete without that! Find a photo, tell Photoshop to add a fish (after moving the arm a little and fixing the background) and YOWZAH!! Shane’s got a fish!

I took the Shane-With-Fish photo into Stable Diffusion for a bit of a face touch-up.

Do I actually look like any of those?

Not one bit of it! But that is the point, innit? Instead of showing my real face, I swapped in the faces of some celebrities on Erin’s hallpass list. But not all of them were for Erin’s viewing pleasure. One reveals me more so than the others. If you can follow the clues. I think that’s a riddle with which even an AI would have difficulty.

Intelligence: There’s an App for That!

And there we have it, a handful of ways that AI platforms can be somewhat useful to swingers. One day there will be a swinger hook-up AI who will comb through the various lifestyle dating sites and tell you which one is the most used in your area. (I haven’t tried to get Bard to do this yet, but it’s so wary about adult content, that I hesitate to assign such a task to it.)

So far, AI technology is not the human-hating, war-mongering death machine sci-fi has led us to believe it would be. I don’t foresee a Skynet situation in our future. Building a Terminator army isn’t how AI is going to destroy us. Nor will we end up as batteries powering the world of the machines. (In an effort to be positive and not give the AIs any ideas, I will refrain from presenting my theory about the most likely way the AI will destroy us. But I do have one.)

We’re already seeing massive shifts in the jobs/careers paths. A couple of people who know how to use AI well can do the work of entire departments. It might help small businesses be more flexible and soluble, which would be great. But there will be huge accessibility issues.

Yes, we can all use Bard and ChatGPT, but not everyone has the time to learn prompt engineering. Even now, some schools are banning the use of AI while others are embracing it. Which students will be prepared for when employers require an adequate understanding of using AI in your field?

Eventually, proprietary AIs that have been specially trained with well-refined datasets will outperform the public-facing, for-free versions. Companies that can afford such machines will once again have a leg up on small businesses that cannot.

We live in interesting times. Try to enjoy them.