New World Notes

I Thought ChatGPT Was Impressive. But Then Someone Asked ChatGPT About a Topic I've Written Two Books On.

 

ChatGPT continues to generate incredible hype, buzz around Silicon Valley, and grand pronouncements that its soon going to replace writers and teachers, even accountants and lawyers.

 

But before you believe any of that, here's a simple test to try:

 

Ask ChatGPT about a specific subject that you know very well. And take a close look at the results.

 

I thought about that when Redditor "FLAPJACKU" asked ChatGPT a simple question: "What can Second Life do to improve its systems?"

 

The answer (see above) reminds me of that famous story about the amazing harpsichord-playing robot from the 19th century:

 

Professors Warn Students About Cheating With ChatGPT With Stern Warning… Created by ChatGPT

You've probably heard about students using ChatGPT to cheat on their essay writing work, and it's something developer Catherine Winters knows about first-hand from colleagues at a local university.

"[They are] freaked out because of ChatGPT -- it looks like there was a lot of cheating the last couple weeks of last term," she tells me. "So they've banned it, added the use of it (and anything the prof doesn't allow) to be academic misconduct, etc."

Catherine came up with an even better solution: Asked ChatGPT to create the ultimate warning against students using ChatGPT. (See above.)

"I know at least two of them have since put that in their syllabi as warnings that they know what ChatGPT is," Catherine tells me. "Which is doubly hilarious." (Educators, the full anti-ChatGPT text is below, for your own copy/pasting needs.)

 

This, plus news of an AI programmed to detect AI, is starting to remind me of the artificial intelligence arms race in Philip K. Dick's classic short story "Second Variety": The AIs on opposing sides kill off most the actual humans in the war, keep on fighting against each other... and then set to work creating new AIs programmed to appear human enough to fool the enemy AI. 

 

So could students could train ChatGPT to generate answers that are designed to disguise their origins as coming from ChatGPT?

LOOK: These Beautiful Paintings Are Actually Second Life Screenshots Processed Though Stable Diffusion AI

Cajsa Lilliehook covers the best in virtual world screenshot art and digital painting

Pavel Stransky is doing something really interesting with AI art. He starts with an SL screenshot, then processes it through Stable Diffusion, a text-to-image Artificial Intelligence, fine tunes it with Photoshop and then adds detail and definition with another AI called GigaPixel. So from an SL screenshot he produces what looks like a Modigliani. Who knows? Perhaps it is Modigliani’s “Seated Nude” as seen from the back. 

For more examples of his SL/AI art, click here:

AI Probably Won't Disrupt 3D World Creation, Argues Veteran Game Artist

Above: The procedurally-generated No Man's Sky

With the growing popularity of AI-driven 2D image generators, it's no surprise that we're starting to see many people speculating about how AI-driven 3D image generators may soon transform games and metaverse platforms. (For instance, well, AM Radio and me.)

However, veteran game industry artist Aura Trilio just hit me with a knowledgeable and refreshing blast of anti-hype, pointing out how difficult (if not impossible) it would be to have an AI which can feasibly create whole 3D worlds:

"AI isn't going to affect any field that doesn't have a giant database of free (stolen) training data for it to absorb," she explained. "There aren't enough 3d models in existence for basically any model to eat and spit out anything usable. Even most bespoke, handmade 3d model generation algorithms spit out models that are completely unusable in games because the logic behind character creation and topology is extremely precise and needs to be carefully thought out. So: it's not going to change it."

But wait: The acclaimed game No Man's Sky generated a whole galaxy of content through automated procedural generation. Doesn't that suggest its feasibility for other games/worlds?

LOOK: This Artist Is Creating an Entire Graphic Novel With Midjourney Credited as Co-Creator

Cajsa Lilliehook covers the best in virtual world screenshot art and digital painting

Zoe Jimenez is primarily a Second Life creator and photographer with an obsession with graphic novels and magazine covers. Her Second Life photos are full of covers and character cards, most not safe for Google. However, with the advent of Midjourney, she took a departure from the virtual world and went into a creative frenzy, refining her skill with the AI until she created “Proof of Concept”, her Midjourney-illustrated graphic novel. You will note at the bottom of the page, she explicitly lists Midjourney as a co-creator, the kind of transparency that AI creation demands. This is page three of her story and it’s hilarious. I love when he says she is too much and she replies, “I will be.” 

For a peek behind the scenes, click here:

Lensa Scraping Artists' Work Without Their Consent Into Its "AI" Images -- Even the Artists' Signatures!

The Lensa app has gone viral in recent weeks, with much excitement over its new AI-driven “magic avatars” feature.

Small problem: It's not exactly magic nor purely artificial intelligence. Instead, to create these avatars, the app is apparently scraping up artists' images without their consent. The image appropriation is so blatant in many cases, the Lensa-generated images even include the original artist's signature. (See example above.)

I learned this via professional illustrator Lauryn Ipusm, who's been posting examples of this on her Twitter in recent days.

 "I think they didn’t think artists would stand up for themselves because we don’t [have] industry labels the way the music industry does," Lauren tells me. She says that, because Stable Diffusion, the AI company which provides Lensa's neural network, is very careful about how the platform samples and trains from recorded music.

"The fact that they do so with their music model shows they are well aware of copyright (I mean it’s a basic concept, anyone who isn’t a little kid is aware of copyright), and that it’s not something that was too complicated to implement."

Lauryn's findings were noted yesterday by NBC News, but so far, Lensa's corporate owner has only offered an incomplete reply:

Artist Uses ChatGPT AI to Create Text-Based Adventure Game -- Then Feeds That Text Into Midjourney AI to Create Images for Game

Game industry artist Martin Nebelong recently started experimenting with ChatGPT, and discovered he was able to quickly create an entire "choose your adventure"-style adventure game through the AI text generator. Taking that a step further, he then took the ChatGPT text and fed it into Midjourney, the AI image generator, which then created illustrations to accompany the game.  (Watch above.)

"I was very surprised by how quickly the chat bot was in on the idea," Martin tells me. "To begin with, it wanted to write the whole story including all branches in one reply. But when I asked it to serve them to me as a story bit with a few branching options, to wait for my selection and only then to move on in the story, it got that an started over with the correct format. When it wanted to wrap up the story faster than I wanted, I just asked it to do a more detailed story and it did so.

"I guess the most impressive part to me was when I asked it to describe the characters in high detail for the prompts I wanted to use in Midjourney. It gave very detailed descriptions of all the characters including their clothing, their look, personality, names etc." (See an example below.)

But he was also impressed at how ChatGPT added descriptive flourishes he wasn't expecting:

LOOK: Artist Blends Real Life Photos of Herself With Her Second Life Avatar in Midjourney to Amazing Effect

Five faces of Whiskey Monday and her IRL owner -- via her Flickr

You can create an avatar based on yourself in Midjourney, but famed Second Life artist Whiskey Monday went a step further, combining multiple images of her SL avatar with photos of herself in real life to create an incredible new persona. 

"I put my RL pics first, and then it generated a Midjourney version of me," Whiskey explains. "Then I added SL pics, and did the same thing, and then I blended them all in several steps. I found that Midjourney doesn't do well with glasses, so I tried an older pic without glasses." (See above.)

In the process, she learned something about Whiskey in SL and Whiskey IRL:

"I used to try hard to make my avatar look like RL me. But over the past couple of years, I’ve stopped caring so much and have just played with new faces for Whiskey. I think it would be interesting to try this again with an older Whiskey face. But what I learned this round is that I still have a certain look, even though Whiskey is less like Jenn." (I.E. Whiskey's offline owner.)

Here's some other variations courtesy Ms. Day:

Artist Feeds Her Childhood Diary Into an AI Chatbot -- And Has a Moving Conversation With Her Younger Self

What you're looking at above is an excerpt from a moving conversation between tech artist Michelle Huang, and Michelle Huang as a young girl.

"I trained an AI chatbot on my childhood journal entries - so that I could engage in real-time dialogue with my 'inner child'," as she recently explained on Twitter:

I kept diaries for about 10+ years of my life, writing almost everyday — about my dreams, fears, secrets... I used GPT-3 as my playground, and ended up taking samples of text from a bunch of different entries that I felt were representative of my personality and values during that time... this way, i could accurately simulate what it would be like to talk to my childhood self, based on real data sources during that time period.

Michelle tells me she learned something crucial about herself by engaging with this simulation of her younger self:

"I would say, for me personally - the kindness and understanding I felt from my 'past self' helped me untangle some knots and stuckness that I felt," as she puts it. "It illuminated to me how hard I was on myself, but how much I extended understanding to others. It felt really rewarding to receive this kindness as well, from a younger version of myself.

"From a broader learnings [perspective], what stood out to me was the ability for technology to be used for mental health and wonder."

I also see amazing potential in this for NPCs and avatars. Imagine entire digital personae you can engage with that are directly based on yourself. Or for that matter, NPCs directly based on novelists, poets, and public speakers from history and fiction.

Michelle agrees on that front:

"This is the stuff I think that has the most interesting ramifications: more broadly, more immersive human / computer interface loops, from conversation with virtual therapists to in-game interactions for virtual worlds, given there is user input, AI could be used to train highly customizable responses or generate unique storylines per use."

It could even become the seed of a new virtual world product:

Midjourney Can Turn Your IRL Photo Into a Supercool Avatar - Here's How

Update, 6:35pm PT: I'm having trouble replicating this process with my own photo. If you have advice/tips, please post in Comments! 

Update 2, 11/18: Finally got this to work on my own photo. The original instruction included brackets and quotation marks but those aren't necessary when communicating with Midjourney.

The photo at top left is Nick Yee, founder of gamer market research firm Quantic Foundry; the others are variations of Nick in different genres, as interpreted by Midjourney. 

"It was always possible to do image-based prompts in Midjourney," Nick explained on social media recently, "but they used to be treated very loosely and couldn't be used to replicate faces. The newest version has much stickier image-based prompting."

Here's how you do that: Find a photo of yourself online, and then enter something like this into Midjourney's image prompt:

/imagine <image link> fantasy-hunter-warrior

/imagine <image link> futuristic-cyberpunk-hacker

/imagine <image link> captain-america

(Or instead of Captain America/cyberpunk/etc., whatever evocative genre terms you want to use, of course.)

The results are pretty impressive, especially since AI image readers have a notorious problem with racial bias.