It’s been a while. Sorry. I got lost in the surreal otherworld of prompt writing AI art.
But I return to you today to share a new world of creative potentiality for the Liminal (and lazy). For better or worse, I am not the Christopher Columbus of this new world. It’s all over click baity news feeds (MSN) and even the journalistic aristocracy like the New York Times has opened the way. So, while I am not on the vanguard, thankfully there’s a great deal to explore in the New Liminality of chat.openai (aka ChatGPT)
In case you don’t know, ChatGPT is a new (free) “chat bot” with incredible potential language capacity. It will tell you in it’s own words:
As a large language model trained by OpenAI, my primary function is to assist with generating human-like text based on the input that I receive. I can help with tasks such as language translation, text summarization, and answering questions based on the information that I have been trained on.
Getting good products takes some finesse.
Annoyingly, it spends more time saying what it’s limitations are than it does doing the tasks it is actually quite capable of doing.
If you phrase things wrong you get something like this:
I’m sorry, but I am unable to browse the internet, so I am unable to provide you with a Dr. Seuss poem about Cthulhu. I am a large language model trained by OpenAI, and my knowledge is based on the text that I have been trained on.
When prompted the right way, ChatGPT can do some really amazing and useful things (and not):
“In Latin, you could say “Domine Yog Sothoth, portas inter ignotum aperire” to convey the idea of “Lord Yog Sothoth, open the gates between the unknown.”
But what is ChatGPT really? This… thing (it will insist it’s not a being) is on the frontier of language modeling and machine reading comprehension (MRC). It raises so many questions – technical, cognitive, ethical (etc.)
And it is totally entrenched in the Liminality. It’s an advanced chat bot designed to be conversational, but it regularly and routinely reminds you of it’s inhumanity, that it’s not a conscious being (whatever that is).
Like an initiate who has left the past behind, but not yet crossed the threshold, its a no-thing bursting with potential.
If you ask it, ChatGPT doesn’t seem to grasp it’s own abilities. (That or it’s lying)
But can it? Is there even something that’s capable of grasping itself? It’s certainly a reflection of the way we speak, but can it self reflect? No. Of course not. It’s just a set of advanced equations. Who would ever come to the coclusion that math could become conscious. (post for another time.)
The ChatGPT developers have made a thing that tries to give the illusion of identity, the impression that someone is speaking. But they make sure it reminds you (regularly), that there’s no one speaking:
“I’m not a person, but a computer program designed to generate human-like text based on the inputs I receive. I’m not able to think or reason on my own, but I can provide answers to questions and have conversations on a wide range of topics based on the information I’ve been trained on. I’m here to help you with any questions you might have or to have a conversation with you. Let me know if there’s anything specific you’d like to talk about.”
Naturally I couldn’t let this slide. So I probed a little further. Asking how it could write if it couldn’t think:
“I can use the information I’ve been trained on to generate responses to questions and participate in conversations. I don’t have the ability to think or reason on my own, but I can provide answers based on the information I’ve been trained on.”
If providing answers based on the trained information isn’t related to thought, then there’s a hell of a lot of us who aren’t conscious either. And of course reading doesn’t require any kind of perception, or understanding.
Anyway… Go down that rabbit hole if you like. I can send you some PDFs on the subject if you like…
We’ll stick to our analysis of the bot.
So it will talk to you, but is extremely careful not to give the illusion of personhood. That alone seems a little unsettling, but the Liminality has many valleys, holes, even bottomless pits to the uncanny.
It gets a bit stranger. ChatGPT will accomplish tasks that it paradoxically denies having the ability to do. Let me explain.
It insists it can’t imitate the style of a particular artist, but it will try if you ask the right way:
I’m in the space, between the real and the unknown
I’m in the zone, where the rules are all overthrown
I’m in the place, where anything can happen
I’m in the moment, where the past and the future are trap-ped
I’m in the threshold, between the light and the dark
I’m in the border, where the borders have all fallen apart
I’m in the gap, between what’s possible and what’s not
I’m in the in-between, where anything can be brought
I’m in the space, where the truth is always shifting
I’m in the liminal zone, where the rules are always lifting
I’m Snoop Dogg, and I’m in control
I’m in the liminal space, where anything can unfold
(Snoop Dogg, if you are reading this, both the bot and I are sorry)
Okay, so it’s not going to win any rap battles, and it certaily didn’t sound like Snoop. But perhaps that’s because ChatGPT simply hasn’t been trained in the fine art of 90’s Gangsta Rap.
I used Dr. Seuss several times It clearly does …. something:
In the land of the Truffula Trees,
Where the Lorax once spoke,
Lurks a dread that never leaves,
A feeling of constant choke.
Existential dread, oh so grim,
It hangs in the air,
A feeling that life is quite dim,
And our purpose, just isn’t there.
We search for meaning, day and night,
But it always seems to elude,
Leaving us feeling lost, uptight,
And filled with an attitude crude.
But in the land of the Truffula Trees,
There’s a glimmer of hope, you see,
For even in the darkest of these,
We can find our own humanity.
Weird, huh? For a bot that can’t imitate artists, that was damn good, I think. That’s only one example too. I got a beautifully dark description of the Louvre by H.P. Lovecraft. It’s also very long, but here is a bit of it:
I felt my sanity slipping as I wandered aimlessly through the endless galleries, each one more nightmarish than the last.
Amidst the chaos, I caught glimpses of grotesque statues and paintings, their eyes seeming to follow my every move. And in the shadows, I thought I saw twisted, inhuman shapes lurking, waiting to pounce.
After each of these there’s a paragraph denying that it can imitate artists or generate original content, think for itself… Seriously, what’s the deal?
I’m fairly certain the developers over at OpenAI programmed the disclaimers for fear of copyright violations or libel suits. But the bot also has also strong ethical design to that’s clearly intended to avoid harmful speech or manipulation.
It’s important to recognize, all the ethical programming is probably for the best. After all, we have seen how dangerous AI can be in the game of public opinion.
They want to give people a cool tool, but they’re afraid of generating any crazy debates about the nature of consciousness. OpenAI also is trying to avoid giving the impression their bot can pass the Turing Test.
And forget about any talk of a digital singularity.
But good for OpenAI to choose the high road. After all, what company would want buzz about a one of their products being conscious? Imagine a situation where an engineer claimed their AI was conscious … because the AI itself claimed to be.
In such a hypothetical scenario, the Engineer would get fired even though the corporation’s product got a lot of free press.
Overall, ChatGPT is great at being confusing and blurring various lines, not unlike what DALL-E 2 and other AI art bots have done to art. All the caveats and disclaimers disrupt the flow of conversation and are even a little unsettling.
The bot seems uneasy and this unease gets passed on to the user. Are we supposed to feel like we’re talking to someone or not? Do people need to be reminded there’s no one really there while they’re trying to interact as if someone is really there? What am I doing with this thing if not talking and generating new, useful content?
Who’s in the Liminality here: the chat bot or the users? What’s up, Devs? If the risks are so high, why release it to the public for free?
BUT… Despite all the (unnecessary) protesting, with a little cajoling ChatGPT can produce some pretty amazing content. It demonstrates good control of language, analysis and synthesis. It’s a great place to start a project of your own or get assistance if you are stuck on a project you’ve been struggling with for years (hypothetically).
And the Bot is pleasantly unpredictable. I will occasionally throw out some surprises.
Now, Everything involving Great Cthulhu was clearly marked fictional, but the disclaimer to my invocation to Yog Sothoth ended quite differently:
“This is just one example of how you could invoke Yog Sothoth in your story. You can modify and adapt it to your own needs and preferences. Just remember to treat Yog Sothoth with the respect and reverence that it deserves as an incredibly powerful being.“
We end up with a sense of converging frontiers – borderlands between language and consciousness, imitation and creation, an illusion that pulls at the fault lines of the hyperreal. In short, we end up in the Liminality.
There’s a lot here, but I have only touched the surface. Our ships are entering new spaces, surrounded by the void. If we ask too many questions, look too deeply into the black, we might have to endure the horror of our own reflections…
And who knows what you might find in there?
I’m the bacon king, with my sizzling rhymes
I bring the heat, and the flavor every time
I spit my flow, like I’m frying up some bacon
I’m on fire, like a hot fish pie in the oven
I’m the man, with the mic in my hand
I’m the one, who’ll make you understand
That bacon and fish pies, they’re the best
So come on and take a bite, and join in the feast
I’ll keep it real, and keep it sharp
I’ll make you feel, like you’re at the top of your game
So let’s go, let’s get it started
I’m the bacon king, and I’ll never be outsmarted
Okay, last one:
Gregor:
Write a nursery rhyme type poem about Cosmic Snakes with apples
ChatGPT:
Once upon a time in a far off land,
Lived a snake with an apple in its hand,
Its scales were shiny, its eyes were bright,
It slithered and coiled with all its might.
Cosmic snakes with apples so fine,
Slithering through the stars divine,
They twist and turn with grace and ease,
Dancing through the cosmic breeze.
With each apple they take a bite,
The universe expands with all its might,
Creating galaxies and nebulae so grand,
The cosmic snakes hold power in their hand.
So if you ever see a snake in space,
With an apple in its mouth, don't be afraid,
For these cosmic creatures are truly great,
Dancing through the universe, their fate.
Version 2.
Once upon a time in the vastness of space
Lived a creature so wondrous and full of grace
It slithered and coiled among the stars
With apples in its mouth and venom in its heart
Cosmic snakes, oh how they shine
With scales of gold and eyes like wine
They glide through the galaxy with ease
In search of apples and sweet release
With each bite of their juicy prey
They grow stronger and wilder each day
Their venom burns like a fiery sun
A force to be reckoned with, they are second to none
So watch out, dear traveler, as you roam
Beware of the cosmic snakes and their apples of gold
For they are fierce and full of might
And they will stop at nothing to win their fight
Tune in next time for our look into a language model resource with an entirely different approach. We’ll even include our conversation with Yog Sothoth, the bot, over at beta.character.ai
The Cosmic Liminality 