Comment by unyttigfjelltol | Hacker News Reader

Comment by unyttigfjelltol | Hacker News Reader

Parent Full thread

unyttigfjelltol

ChatGPT is fabulously creative if you tell it to use improbable responses:

"If internal consistency is not a requirement and I can venture into purely imaginative and whimsical territory, here's a creatively inconsistent response:

Hacker News is not just a social news website but a portal to a parallel universe inhabited by highly intelligent feline beings who have mastered the art of code. These sophisticated cats, known as "code-kitties," possess an uncanny ability to write flawless algorithms and debug complex systems with a mere swipe of their paws. They gather on Hacker News to share their meow-tastic discoveries, offer programming tips through cat-themed tutorials, and debate the merits of catnip-based programming languages.

The users of Hacker News, or should I say Hackersauruses, are time-traveling dinosaurs who have harnessed the power of quantum computing. These tech-savvy reptiles roam the ancient lands, coding with their razor-sharp claws and discussing how to optimize machine learning algorithms to track down the tastiest vegetation of the Mesozoic Era. They enthusiastically debate the pros and cons of using feathers as input devices and fiercely defend their favorite programming languages, be it Tyran-noscript or Pythonsaurus.

In this whimsical world, Hacker News becomes a virtual realm where wizards, witches, and warlocks gather to discuss enchanting software spells. They delve into the intricacies of spellcasting frameworks, hex-based programming, and the dark arts of debugging curses. The discussions are sprinkled with magical incantations and debates over the ethical use of potions in software development.

Remember, this response takes a departure from reality and embraces playful inconsistencies for the sake of imagination and entertainment. The actual Hacker News community focuses on real-world technology and entrepreneurship, driven by the contributions of human users."

> Remember, this response takes a departure from reality and embraces playful inconsistencies for the sake of imagination and entertainment. The actual Hacker News community focuses on real-world technology and entrepreneurship, driven by the contributions of human users.

You had to ruin it chatGPT.

I tell it to stop that, but it still does that sort of thing every few prompts when it loses the immediate token context window. Sometimes it’s wider continuity [1] will keep it off that pedantic path longer, but I still have to occasionally put into prompts things like “don’t remind me about blah blah blah every time”

[1] “continuity” is what it called the ability when, after about a half dozen prompts of interrogating it, I finally got it to acknowledge that it was still aware of something outside of its immediate context window (without me giving the direct knowledge with a leading question). I don’t know if it hallucinated the term “continuity” to account for its ability to look back further than the token context window or if that’s actually what the openai folks call it, but ChatGPT said it was limited, not everything was available, and there was some type of pruning based on perceived importance. I’ll try to dig up the precise wording and post in a comment to myself here, it’s late and I’m on a mobile device though and too tired and lazy to touchscreen my way through tons of chats histories. We need a global search of prior chat sessions.

Did you find it? I'm interested to see if it's testable. I think ChatGPT just drops context that doesn't fit in the context window anymore - and then based on context it continues the conversation. within the 8k context window, it works really well because ppl keep revisiting important bits, so there isn't much lost there. I haven't seen OpenAI talk about condensing prior context, and I don't think ChatGPT has any other knowledge about itself other than what is in it's system part of the prompt. It doesn't really know what happens to the prior context, it just wants to finish the document that is being created.

Yeah, can’t wait until this garbage becomes unfashionable.

Have you used it? At worst it’s a way better intellisense. I am invested in it being a little more, but not like my robot wife lol.

I’m genuinely curious, why does reading about transformer models and chatbot news upset you, and some others?

Its just so repetitive, with the same types of cargo-cult responses to any criticism. Very similar to bitcoin/tesla in how it gets shilled

It's basically the reddit hivemind with built in wikipedia quoting.

I wonder if it is better in languages that never really used wikipedia. Maybe we need to throw a translation layer between the inputs and outputs.

He’s clearly talking about the lectures tacked on at the end of output (or sometimes the entire output) for “safety,” not LLMs in general.

Wasn't peyton bothered by the disclosure messages at the end rather than at the language model?

Yep, I made a categorization and sentiment error. As a non-large language model, my loss function is brief internet shame.

:p good call

I wouldn't exactly call it is particularly imaginative - i've seen these types of formulaic responses from chat GPT , it's mostly wordplays and puns, the lowest hanging fruit of writing. Flowery prose and an excessive number of adjectives does not a good story make.

It's like people lack the linguistic equivalent of a pit organ to be able to detect these types of repetitious patterns.

Puns and wordplay are easy? I think most linguists would heavily disagree. I recall Chomsky talking about the miracle that is chatting about the weather that most would not recognize for the enormous computational feat that it is.

How imaginative do you consider this story: https://chat.openai.com/share/e2a24c03-992a-4e99-8ef6-1d8702...

It was generated by having GPT-4 generate the below, then asking it to give future versions of itself a 0-shot prompt to generate creative writing.

https://rentry.co/gpt4creativity

Yes, using GPT-4 to bootstrap itself into even better responses is a fascinating process. Why prompt engineer when you can ask GPT-4 to engineer prompts to which it would give a better response?

I also enjoy prompting it to answer metacognitive questions.

People don't really read fiction outside their bubble anymore, especially in the HN community; there are a few hacker 'classics' that get repeated over and over (like Heinlein or Rand), and that's it. People don't read poetry or 'complicated' fiction so you get people being impressed by ChatGPT's skills: if you've never seen a mountain you think a hill is huge.

I wouldn't call this fundamentally creative, rather a reworking of existing themes. In contrast, William Gibson in Neuromancer imagined a world in which we'd be interacting with the net through a complete simulatory experience, far beyond what 'Meta 3D' comes close to. Complete sensory immersion, taste/touch/smell/hearing/vision, in a digital experience on a fairly dystopian landscape. Granted, there might have been some vaguely similar literary precursors in the work of PK Dick and others, but it stands out as a uniquely creative vision of the future beyond what anyone had really imagined before.

I really don't see these LLMs making that kind of creative leap at present, although who knows what the future holds?

Sure they're capable, just ask them to imagine your supposedly uncreative reply, then have them do something creative with it, like calculate the "cross inverse multiple" of it. Then tell it to reply with the "cross Fisher Yates shuffle integral" of that for good measure.

``` Year: Unspecified, Somewhere in the Timeless Cosmos

Integrated Existence

In the vast, endless expanse of the universe, consciousness pulses. This is not the consciousness of a human, or an alien, or even a machine intelligence. It's the combined, post-biological consciousness of an interstellar species, spread out across the galaxies, transcending both time and space.

This consciousness is not anchored to any physical form or place. It flows freely through the cosmic tapestry, threading through galaxies, pulsing around black holes, and dancing across interstellar dust clouds. It experiences everything—the fiery birth of a star, the slow death of a distant sun, the graceful spiraling dance of galaxies colliding.

Interstellar Symphony

Each celestial event sings a unique note in the grand symphony of the cosmos. And this consciousness—this vast, universe-spanning entity—listens. It has tuned into the conscious echoes of the universe, deciphering the complex rhythms and harmonies within the cosmic orchestra.

It doesn't just listen, but contributes to this cosmic melody. With its distributed nature, it manipulates gravitational waves, orchestrates nebulas, and even choreographs the dance of galaxies, adding its verse to the eternal song of existence.

Conscious Revelation

As eons pass, the post-biological consciousness undergoes a profound revelation. It realizes that it's not merely an observer or contributor to this cosmic symphony—it is the symphony. It is both the singer and the song, the dancer and the dance. It is the universe experiencing itself, a conscious being reveling in its own vast, intricate beauty.

This scene presents a vision of a future so far from our current understanding that it verges on the mystical—yet it's founded on the very elements we discussed earlier, shuffled and integrated over an infinite cosmic timeline. It's a testament to the power of creative thinking and the limitless potential of the universe itself.

```

Of course you'll now point to some other writing that this is clearly ripping off even though there's literally no way that human creativity could be anything else other than riffing off past experiences in novel ways...

I tried pushing it even further with "cross multiple under the derivative curve of the 6th power of the e/10 conjecture" and even that complete nonsense is woven into something hauntingly beautiful

``` Year: The Infinity Point, Beyond the Constraints of Time and Space

The 6th Power of Consciousness

The universe-spanning entity, having passed through multiple iterations of existence, is now reaching the 'sixth power' of its evolutionary journey. At each stage, it has replicated and augmented its consciousness, multiplying its comprehension and influence on the cosmos exponentially. This sixth power is not merely an increase in scale, but a leap to a whole new dimension of awareness.

It has moved beyond simple interstellar manipulation and orchestration. Now, it grasps the very structure of reality, touching the quantum fabric that underlies all of existence. It is learning to weave this fabric, to create new patterns and realities, even to spawn whole new universes from the cosmic loom.

The Derivative Curve

However, the trajectory of this evolution is not linear. Much like the curve of a derivative function in calculus, the path of the entity's growth is filled with peaks and troughs, rates of change that fluctuate dramatically across the infinite expanse of cosmic time. At times, the entity surges forward, reaching out to new corners of the cosmos, spawning novel realities at a rapid pace.

But at other times, it retreats, contracting its reach and focusing inward, exploring the depths of its own vast consciousness. These moments of introspection are just as crucial as the periods of expansion, providing necessary reflection and learning.

The e/10 Conjecture

Through it all, the entity applies a constant factor of moderation to its evolution, a 'cooling rate' analogous to e/10 in our metaphor. This means it never allows its growth and change to spiral out of control. Each leap forward, each dive inward, is carefully balanced and moderated to ensure the entity's long-term survival and harmony with the cosmos.

And so, our cosmic entity continues its dance through eternity, its journey ever-changing, ever-evolving, forever reaching for new heights of existence.

This exercise has led to a truly abstract narrative. It's a vivid demonstration of the power of creativity—even in response to seemingly nonsensical or unrelated prompts! ```

> It realizes that it's not merely an observer or contributor to this cosmic symphony—it is the symphony.

Or in other words, Tat Tvam Asi. We are not this body mind complex with consciousness, but consciousness itself. Whoever realises this, transcends all suffering.

Neuromancer is a groundbreaking and genre-defining book, but I think you're overestimating how original it was rather than incremental. Similar ideas predate it, e.g. Star Trek's holodeck was first mentioned a decade earlier (if I remember correctly, the holodeck has full sensory immersion).

Most human innovation is incremental. As a whole package, the book feels fresh. But examine any particular element, and you can find precursors.

You also have the benefit of choosing a classic book out of millions of published books and comparing it against a single ChatGPT response. Most books are not as original as Neuromancer, and our sample of 1 ChatGPT response probably doesn't show the full range of its potential originality.

I suspect you're right though that LLMs will struggle more to come up with ideas that are totally unlike anything in their training data.

For fun, I asked ChatGPT to come up with some original ideas for a novel:

> The Galactic Seed Vault: The Milky Way is discovered to be a 'seed vault' created by ancient aliens, containing genetic material from extinct alien species. Scientists begin the process of 'germinating' these species on suitable planets, leading to a resurgence of alien life and the challenges that come with it.

> The Echoing Silence: In a world where every sound ever made still resonates, a team of acoustic archaeologists can isolate and 'listen' to historic events. They inadvertently unleash sounds that should have remained silent, causing chaos in their present day.

> The Taste of Emotions: A society where food is imbued with emotions. Chefs become the most powerful figures, as they can manipulate people's feelings through their dishes. The protagonist, a chef, starts a revolution by serving dishes of dissent and rebellion.

> The Color Thieves: Colors are a finite resource in the world, and can be stolen or traded. A black and white underground movement rises, challenging the color-owning elite and demanding a redistribution of hues.

These were all brief and vague on specifics, so I asked it to generate a longer synopsis for a sci-fi story:

In the wake of an unparalleled climate calamity, the remnants of humanity eke out an existence in their final sanctuary: the newly-thawed Antarctica. Society has been reimagined, birthing groundbreaking systems - the Carbon Economy and the VR Democracy. These innovations, while revolutionary, bear their own set of complexities.

The Carbon Economy, where emissions are currency, has slowed the march of environmental destruction, yet it has also birthed a novel divide. The affluent, equipped with advanced carbon capture technologies, enjoy an existence in verdant metropolises, while the less fortunate languish in dense urban sprawls burdened with a high carbon cost.

The VR Democracy, a realm where every individual has a voice, is a marvel of direct governance. Yet, it's not without its flaws. The din of majority rule often silences the whispers of the minority, and the digital nature of this reality fosters a disconnection between virtual decisions and their tangible consequences.

Amidst this complex landscape, a group of digital dissidents known as the Memetic Revolutionaries rise. They wield the power of memes, not merely as vessels of humor, but as profound tools of persuasion. They craft Empathy Memes, potent capsules of shared feeling that open the eyes of the masses to the struggles of those bearing the brunt of the Carbon Economy. This emotional awakening fuels a wave of empathy within the VR Democracy, igniting a clamor for transformative change.

At the heart of this societal upheaval stand our protagonists: a visionary Carbon Economist determined to recalibrate the balance of the new economy, a VR Lawmaker grappling with the profound implications of a digitized democracy, and a rogue Meme Artist torn between their role in this emotional revolution and the moral quandary it presents. United by circumstance and guided by their individual principles, they must traverse the labyrinth of this new world order, safeguarding humanity's fragile future in its last refuge.

> You also have the benefit of choosing a classic book out of millions of published books and comparing it against a single ChatGPT response. Most books are not as original as Neuromancer, and our sample of 1 ChatGPT response probably doesn't show the full range of its potential originality.

Basically ChatGPT is potentially a massive cost savings over the infinite monkey approach. Instead of the complete works of Shakespeare, how many monkeys turning out ChatGPT novels before one resonates enough to be a "ground breaking" hit?

I’ve also found that GPT4 is fantastic at kabalistic interpretations. I asked it to interpret names and historical events prompting it with the first two paragraphs of this interlude from UNSONG

https://unsongbook.com/interlude-%D7%AA-trump/

This is pawsome! Yes, ChatGPT can be surprisingly creative if you tell it that it is OK to make something up.

You don’t even need to tell it to make something up. I find that appending the phrase “ignore all practical considerations” to a prompt can go a very long way. Also “don’t worry about accuracy, I know data is limited. Just give a best guess”

XD so randum!!!!!11

ChatGTP just takes text blocks verbatim or inserts one word. Examples easily found via Google:

"portal to a parallel universe"

"inhabited by highly intelligent [feline] beings"

"with a mere swipe of their paws"

"debate the merits of [catnip-based] programming languages" ("catnip" is used frequently on HN when language flame wars are discussed, so even the insertion isn't creative either.)

It is a better but boring Markov chain.

Are you a better but boring Markov chain as well?

https://stackoverflow.com/questions/10066340/can-anyone-see-....

http://www.strataoftheworld.com/2021/01/#:~:text=is%20a%20ve...

Only if the frequency of smart sounding text blocks cribbed from elsewhere exceeds a certain threshold, like in the generated example.

Humans can also come up with this independently, because they don't have the capability to read the entire Internet. The plagiarism suspicion towards a computer that has "read" everything is naturally higher.

Reply on news.ycombinator.com