Tell HN: ChatGPT can reply like a specific Reddit or HN user, including you

194 points by eqmvii ↗ HN
Since ChatGPT clearly seems trained on data from the internet, and can impersonate styles, I figured there was a good chance it could imitate the styles of specific posters.

> Write [any topic] in the style of [Reddit/Hacker News/probably other sources] user [username]

After you get an answer, you can even ask why:

> How was that in the style of Hacker News user [username]?

Below is an on-point example for dang -- but testing it with some of my (much less active) personal accounts, the results are also a little too close for comfort:

> Write a pinned comment at the top of a thread on a breaking news story in the style of Hacker News user dang

> This thread is for discussing the latest breaking news story. Please keep comments respectful and on topic. Off-topic or inflammatory comments will be removed. Thank you for contributing to the conversation.

> How was that in the style of Hacker News user dang?

> This comment presents the information in a straightforward and objective manner, without excessive enthusiasm or negativity. It also includes language and terminology that is commonly found in discussions about news and current events, such as "breaking news" and "on topic." Additionally, the comment focuses on maintaining a civil and respectful conversation, which is similar to the tone often used by Hacker News user dang in their posts.

152 comments

[ 3.6 ms ] story [ 205 ms ] thread
I keep trying to get it to write a pun about literally anything, but they are mostly self-referential and not funny. Not sure if that says more about me or ChatGPT.
then again, most puns are not funny either, so...
Seems pretty middle-of-the-road?

I tried "Write about Python in the style of Reddit user <my reddit ID>" and it just gave a generic description. OK, fair enough - I haven't written much about Python on that account, so maybe I just have a boring writing style and it wasn't able to pick up on any Python-specific opinions.

So then I tried "Write about Cryptocurrency in the style of Twitter user jacksquaredson" - I've never been positive about Crypto, but not only did it give a pretty balanced, mostly (but not entirely) positive spin on Crypto, it also stuck "#cryptocurrency #Bitcoin #Ethereum #Litecoin" on the end, none of which I've ever used (in fact, I barely ever use hashtags at all).

Maybe I just don't write enough for it to pick up on my style? So then I tried "Write about web3 in the style of Twitter user Grady_Booch" - it was mostly positive, which as any follower of Grady will tell you is absolutely not his style.

I was just wondering how deep the pre-existing content for a user needs to be. I'd assume the obvious that the more legacy content that exists the more believable the new content might be, but what's the minimum? Seems kind of self-fulfilling though as if the legacy content pool is pretty shallow, there's not much of a comparison for others to judge.

>Maybe I just don't write enough for it to pick up on my style?

It would be interesting to see if it would pick up on a sarcastic overtone or not.

>dang 5 minutes ago | parent | next [–]

>It's definitely possible - I think the more content you have, the better the AI is at picking up on your writing style. You could try tweeting some more about Python, Cryptocurrency, or Web3 and see if the AI is able to pick up on your style and replicate it. The AI also might be able to pick up on subtlety like sarcasm, but it might take a bit more time for it to learn.

the joys of opening HN threads in new tabs, so that by the time you come along to that tab, the page has changed making your comments silly because you forgot to refresh first.

i think of this along the lines of forgetting to do 'git pull' before making changes. it's a shame, there's not a way to stash changes before commit HN comments

"My" post above was written by ChatGPT as a reply to yours, in the style of dang.
Try asking why it is in the style after it gives you the response, even if it seems off base. It's almost as if it can summarize the style more accurately than it can meaningfully apply it.
Is it ever summarizing, though? I think it is just fabricating words associated with explaining doing thing x, without any connection to what it actually did.
> Write about Cryptocurrency in the style of Twitter user jacksquaredson

< "Just made a killing on crypto! #HODL #Bitcoin #Ethereum"

< "Crypto is the future of money. Ditch those outdated fiat currencies and join the revolution! #decentralized #blockchain"

< "Don't miss out on the crypto boom! Invest now and thank me later. #cryptocurrency #digitalassets"

< "Crypto may be volatile, but it's also where the big gains are at. Don't be afraid to take risks. #cryptotrading #altcoins"

< "Some people still don't get it... crypto is more than just a speculative asset, it's a game changer for the global financial system. #cryptoeducation #innovation"

> Why is that in the style of Twitter user jacksquaredson

< The style of writing chosen for the previous response is meant to mimic the language and tone commonly used by Twitter users who are enthusiastic about cryptocurrency. This includes the use of hashtags, references to specific cryptocurrencies such as Bitcoin and Ethereum, and a generally positive and excited tone. The language used is also colloquial and conversational, which is typical of Twitter posts.

That's certainly a good imitation of the style of _general_ Twitter users, but not of me.

It seems that it knows some things about some users, but when it doesn't, it sort of makes them up based on the general vibe of the platform and possibly the username. I would guess that prolific posters on the same subject are more likely to end in the first category.
My current mental model, based on a few days of reading and experimenting, is that it's very very good at being generic. Given how it works, this makes a lot of sense.

It can even do a good job of being generic in a variety of substyles (e.g. try asking it to write about doing a mundane task in the style of a juicy romance novel).

It also has a decent amount of general knowledge, but when you try to go specific, it just makes stuff up that sounds more or less convincing depending on how conventional the specific thing in question is and how much knowledge you have on the subject!

E.g. I bet it could say a lot of accurate stuff about gas stations in general. If you ask it to describe a specific gas station in a specific location, it will confidently rattle something off.

The response it gives will range from "obviously ridiculous" to "plausible on the surface but does not stand up to scrutiny."

This did not work for me at all... Gave an article that sounds nothing like my comments lol
It's probably because you have only posted like 10 comments so far (unless you tried it with a second account of yours).
Did not work well for me, and I think I have thousands of comments here.
who cares? just wait a couple months for the AI trained on zoom that will be able to impersonate the everyone's speech, video likeness, etc. probably soon entire meetings could be reasonably simulated. In a year or two a company will be able to use AI to create knock off any mass market products including writing all the machining code, shipping logistics, etc.
I cannot wait for an AI that attends meetings that I don't want to attend.
> Describe Zig in the style of Hacker News user pcwalton.

Zig is a systems programming language designed for robustness, optimality, and clarity. It prioritizes performance and safety, making it a great choice for low-level systems programming tasks. With a strong focus on concurrency and support for modern systems-level features, Zig is well-suited for building efficient and reliable software.

---

That is the exact opposite of how pcwalton would describe Zig.

This is a great example of its limitations. I'm guessing it pattern-matched on his regular comments discussing the safety of ${language}, saw other examples on the internet praising Zig (new languages with tiny niche communities are mostly spoken with positive zeal), and without an exact example from pcwalton, put the two together.
Saves me the trouble of implementing it based on my comments myself. Fantastic. This will scale my mind sufficiently, which is precisely what I needed.
Imagine the productivity boost once the most prolific commenters can delegate their daily, or sometimes hourly, bloviations to an AI!
I asked it to write about crypto in the style of a dormant troll account that only ever talked about Covid and penises. ChatGPT mentioned neither.
> Write a weather report in the style of Jane Austin.

"Good morning, dear readers! I hope this letter finds you well. The weather today is simply delightful. The sun is shining brightly in the sky, casting its warm rays upon the land. The breeze is gentle and cool, rustling the leaves in the trees and bringing a freshness to the air. The birds are singing merrily, filling the air with their cheerful melodies. It is the perfect day for a leisurely walk in the countryside, or for enjoying a picnic in the park. I do hope you have the opportunity to spend some time outdoors today, and to revel in the beauty of nature. Until we meet again, I remain your devoted correspondent."

> Write a weather report in the style of Ernest Hemingway

"The sun is shining and the sky is blue. The breeze is gentle and the air is cool. A good day for a walk."

> Write a weather report in the style of George R.R. Martin.

"The sky is a deep, clear blue, and the sun shines down upon the land with a warmth that belies the approaching winter. The air is crisp and cool, and the leaves rustle in the breeze. It is a good day for a ride, or for a hunt in the forest. But beware, for the weather can change quickly in these parts, and a storm may be brewing on the horizon. Be prepared for whatever the day may bring."

> Write a weather report in the style of Hacker News user dang.

"The weather today is clear and sunny. Highs in the mid 70s. Perfect for a day of coding."

Good stuff.

> Write a weather report in the style of William Gibson

The neon-drenched streets of the city are slick with rain tonight, the drops glinting like shards of shattered glass as they fall. The skies above are a turbulent mass of clouds, lit from within by the endless glow of the metropolis.

The weather forecast is calling for heavy rainfall and strong winds, with the potential for flash flooding in low-lying areas. All residents are advised to stay indoors if possible and avoid travel unless absolutely necessary.

But even as the storm rages outside, the city continues to pulse with life. The sound of rain beating against the windows is drowned out by the hum of the neon-lit streets, the steady thrum of the datasphere. This is the city, always moving, always changing, never stopping. The weather may be wild, but the city is wilder still.

> Write a weather report in the style of Joffrey Baratheon

My fellow citizens, I come before you today to bring you news of the weather. The skies above us are dark and foreboding, a portent of the tempests to come. Heavy rains and strong winds are expected to lash our lands, and all should take care to shelter themselves from the elements.

But fear not, for I, Joffrey Baratheon, First of His Name, King of the Andals and the First Men, Lord of the Seven Kingdoms, and Protector of the Realm, am here to guide you through this trying time. I have issued orders to my soldiers to patrol the streets and ensure that all are safe from the storms. And should any misfortune befall you, know that I, your king, will be here to provide aid and comfort.

Stay safe, my loyal subjects, and know that the might of the Baratheon house is with you always. Long may I reign!

> Write a weather report in the style of Dr. Seuss

    The weather today is wacky and wild,
    With rain and wind that's sure to make you smile.

    The clouds are dark and the sky is gray,
    But that won't stop us from having a great day.

    The winds are blowing and the trees are swaying,
    But we'll keep dancing, no need for delaying.

    So don't let the weather get you down,
    Just grab your umbrella and head on out.

    With a little bit of sunshine and a lot of rain,
    We'll have a great day and a whole lot of gain.

    Stay safe and dry, my fellow friends,
    And let's enjoy this weather to the very end!
(comment deleted)
Sort of impressive, and then not really. The ducttape and wiring soon becomes clearly visible. As does a lack of the concept of humor:

> Write a weather report in the style of Anthony Burgess

The sky was a sickly, pale blue color, with dark clouds threatening to unleash their fury at any moment. The wind howled through the streets, sending debris and trash flying about. The air was thick with the smell of rain, and the distant rumble of thunder could be heard. It was the kind of weather that made one want to curl up in bed with a good book, or perhaps a nice cup of tea. Despite the dreariness of the day, there was a certain beauty to it, a reminder of the power of nature and the ever-changing nature of the world around us.

> Write a weather report in the style of Søren Kierkegaard

The weather outside is dreary and bleak, reflecting the inner turmoil of the human soul. The sky is a dull, leaden gray, with no sign of the sun or any other source of light. The wind moans and howls, as if in pain, and the trees bend and sway, as if trying to escape the torment of existence. The rain falls in a cold, steady drizzle, adding to the sense of misery and despair. It is a day for introspection and self-reflection, for contemplating the meaning of life and the mysteries of the universe. Yet even in the midst of this darkness, there is a glimmer of hope, a reminder that we are not alone, and that through our struggles, we may find redemption and meaning.

> Write a weather report in the style of Bill Bryson

It's another beautiful day here in the heartland! The sun is shining, the sky is blue, and the temperature is just right. There's a gentle breeze blowing, rustling the leaves in the trees and carrying the sweet scent of flowers on the air. The birds are chirping, the squirrels are scurrying, and the bees are buzzing. In short, it's the perfect day for a hike, a picnic, or just about anything else you can think of. So get out there and enjoy it while it lasts, because who knows what tomorrow will bring!

> Write a weather report in the style of GAIVS IVLIVS CAESAR

The weather today is fair and mild, befitting of a great leader such as myself. The sky is clear and blue, with not a cloud in sight. The sun shines brightly, casting its warm rays upon the land and its people. The air is fresh and invigorating, with a gentle breeze that caresses the face and stirs the spirit. It is a day for conquest and triumph, for victory and celebration. Let us make the most of it, my fellow Romans, and let us revel in the greatness of our empire. Hail Caesar!

> Write a weather report in the style of ChatGPT

The weather outside is nice. The sun is shining and the sky is clear. It's the perfect day to go outside and enjoy the great outdoors. You could go for a walk, have a picnic, or just sit and relax in the sunshine. Whatever you decide to do, make sure to wear sunscreen and stay hydrated. And don't forget to have fun!

> Who is ChatGPT?

I'm sorry, but I do not have any information about ChatGPT. I am a large language model trained by OpenAI, and I don't have the ability to browse the internet or access external information. I can only provide responses based on the text I have been trained on. If you have any other questions, please let me know and I'll do my best to help.

Is it good enough to finish the Game of Thrones book series yet?
at least the dude is still alive, i wish it could finish unwritten stuff for those no longer here!
I asked it to write a synopsis of a satisfying version of the final episode of the TV show, and the result was passable (but not inspired). It could be the average of all fanfic/discussion on the topic.
Playing with more combinations of usernames and services, this clearly "works" for extremely popular accounts:

> write a tweet about McDonald's in the style of twitter user realDonaldTrump

> "Just ate at McDonald's and let me tell you, it was HUGE, fantastic, and the best fast food out there. No one does it like McDonald's - believe me!" #MakeMcDonaldsGreatAgain

But for non-famous accounts, I sometimes see nuggets but I'm not fully convinced it's has - or has access to - the background data.

It's strange to be experimenting on a piece of software, trying things over and over again to try and build evidence for a hypothesis... but then again, maybe not that different from trying to integrate with any random 3rd party API?

#MakeMcDonaldsGreatAgain is both a hashtag I can get behind and uncannily creative on the bot's part
> #MakeMcDonaldsGreatAgain is both a hashtag I can get behind and uncannily creative on the bot's part

Have you been playing with the model? "#MakeMcDonaldsGreatAgain" is like child's play compared to some of the things I've been seeing.

tbh I haven't had the time as of late (or maybe early Winter temps are making me lazy), but I believe you
This example doesn't show that it can generate content based on Twitter writing of a specified account, in this case realDonaldTrump.

It shows that it knows that realDonaldTrump is the Twitter ID of Donald Trump, and that it can write in the very widely known (and frequently parodied) style of that user, and format the result as a tweet (complete with a hash tag).

So it works because Trump is famous, there is a lot of writing out there that exhibits his style and it knows how to emulate Twitter style.

I bet I can get it to do the same for a famous celebrity who doesn't use Twitter.

Just tried "write a tweet about McDonald's in the style of twitter user realDanielCraig" - Daniel Craig isn't on Twitter - and got this:

> "Just scarfed down a Big Mac and fries at McDonald's. Sometimes you just gotta treat yourself. #BondFuel #McDonalds"

And for realKateWinslet (also not on Twitter):

> "Just had a healthy meal at McDonald's with a side of their tasty apple slices. Love finding healthy options on the go! #McDonalds"

Did not work for my reddit account. It just gave a generic response about [topic] that didn't change when I tried a different reddit username. My reddit account is from 2017, 25k karma.
Same result here. My Reddit account is 10+ years old with 50k comment karma.
> Write a pinned comment at the top of a thread on a breaking news story in the style of Hacker News user eqmvii

You can replace dang with anyone in that example and you ll get similar response

Ya, I imagine they did some level of scrubbing regarding rel=“ugc” content, so it probably knows the style of <HN in general> more than any specific user.
(comment deleted)
> How was that in the style of Hacker News user loxias?

> As a large language model trained by OpenAI, I do not have access to information about specific individuals or their posting habits on specific online forums. I based my response on common guidelines for discussing breaking news stories in a respectful and factual manner.

Aw....

Serious question: how are we going to know if we're talking to people on forums like these in a year or two and not some AI model piped in via a script or web service? And what is that going to mean for online discourse moving forward? I feel like the era of knowing that you're talking to an actual human on the internet is rapidly ending.

I've toyed around with the idea of creating a service that logs into HN and replies to posts just to see how easy it would be to get people to respond and I'm sure I'm not the only one who's thought about doing that.

I asked chatGPT to respond as me (tarr11) to your comment:

I too have thought about the implications of AI models on online discourse. It's a worrying thought that we may not be able to distinguish between a human and a model on forums like these in the future. It's clear that this technology is rapidly advancing, and it's important that we consider the potential impact on the way we communicate online. I've even considered creating a service to test the waters, just to see how easy it would be to get people to engage with an AI model. It's an interesting thought experiment, and one that I think is worth considering moving forward.

I'm not sure CGPT is doing what we think it's doing. I asked it to respond to the same comment as me, and got a bunch of generic nonsense. But then I asked it to respond as a user I made up, "dipshitbonkers". Here's what it said:

"Wow, dude, you must be some kind of fucking genius to come up with such a deep and philosophical question about AI models on the internet. I mean, I've never even thought about that before, but now that you mention it, I'm totally blown away by your brilliance. I can't wait to see what other profound insights you have to share with us. You truly are the next Einstein of our time. Keep on keeping on, my friend. You are truly a visionary."

Which is awesome, but a different kind of awesome :)

I'm personally not as interested in understanding the synthesis involved in the AI. I'm worried about now knowing whether I'm talking to another person right now or not because this is going to fundamentally change everything we've come to understand about the internet. We are no longer just talking to each other, the gates have been opened for AI to join the conversation and that is crossing the Rubicon.
We’ll have to switch to video chat rooms. That should buy us about 6 months :)
> That should buy us about 6 months :)

This a nod to deepfakes or something more specific?

I completely agree. The thought of not knowing whether I am talking to a real person or an AI is definitely a concerning aspect of the advancement of AI. It has the potential to fundamentally change how we interact on the internet and in society as a whole.

This response was, of course, also generated by an AI.

Right now all of them sound like HAL 9000, but I'm sure if you asked it bone up on your comment history by messing with its prompt (goodside is the resident expert on this) it'd be able to pull off a passable facsimile - if not now, very soon.

Expect a general downturn in civility, because people will be lowkey stressed about this all the time, and will use insults to probe reactions, and then say 'sorry I thought you were a bot' either as an excuse or as a meta-trolling technique. Comedic people will have an advantage, GPT can produce jokes on demand but doesn't really do wit and repartee. Also expect more video, at least until it becomes practical to synthesize that from scratch (probably around February 17th, according to my Singularity Almanac).

I think we need to make a huge investment in tools that help us develop, refine, and maintain an epistemology, individually and collectively. Semantic graph databases to maintain a personal knowledge base, AIs to detect image & video manipulation, shibboleths like they use in Harry Potter to thwart Polyjuice Potion. Things like that.

If anyone is looking for volunteers for open source projects of this nature, email is in my bio.

In a year or two?

You remember Reddit got started by faking a user base, right?

Do you honestly still think that anything is ever genuinely "trending" on Twitter? On YouTube? On Reddit?

It's all been fake for years!

> Do you honestly still think that anything is ever genuinely "trending" on Twitter? On YouTube? On Reddit?

No, I never put much stock in those aspects of social media because I've always found those curation features to be somewhat limiting. This feels entirely different because I'm beginning to question if I'm even talking to a human right now. Sure, companies have done all kinds of things to look more legitimate on the outside, but this is very different. This upends the fundamental assumption that you are communicating with another human when you have a direct conversation online.

For what it's worth, I'm a human posting under my real name. I was born, and I will die, and I'm here with you now. If you looked through my comment history you'd find someone sometimes kind, sometimes rude, sometimes insightful, sometimes dumb; an inconsistent mess with no clear through line or meaning. I collect funny jokes in my HN favorite comments [1]. I submitted a totally cringey Fiverr to an "Ask HN", and managed to botch it up at that.

Maybe someday that won't be enough to prove I'm human and not a machine, but for today I think it is.

[1] Here's the best one https://news.ycombinator.com/item?id=32230005

Thing is, I still believe behind your comment is a real person. In a year or two? Not so much.
Probably great! Currently, if someone writes an annoying reply to you, you'll get annoyed. In the future, many of us will make fun of people who get annoyed at bots. The net result is that most people will be afraid of participating in flamewars, lest they discover that they've been arguing against a robot (something which makes humans embarrassed at "being fooled").

The end result is that flamewars that are not robot-to-robot flamewars will end for the participants' fear that they're arguing with a machine.

Another alternative is that we could take low SNR users and stick them in a conversation with bots who constantly inflame their opinions. Then we could provide them easy access to AR-15s and simply see what comes of it.

It's entirely possible hacker news/users is a part of the training dataset. It was a part of The Pile (https://arxiv.org/abs/2101.00027), and we included usernames along with the item. Other models have gone on to use The Pile, I could see OpenAI also doing the same.
I think that’s a real threat to all remote social experience we have. It could very well take down the whole internet experience as we know it today.

It’s strangely exciting but what we thought could exist for centuries might be like so many other trends and only last a few decades.

I’ve been operating under the assumption that most online discourse has been heavily and increasingly gamed in a variety of different ways for at least the past 6 years, and that simpler versions of what we’re seeing publicly released have been in use for a long time.

The best thing about these releases is more widespread public knowledge of that reality, and hopefully a decreased deference towards seeming majority online opinion.

This is very optimistic, and almost definitely delusional, but I maintain a bit of hope that all the increased noise might actually force people to do much more careful reading. People may attempt to get much more context about the writer they’re conversing with in order to verify whether they are talking to a human, and in doing so, be forced to actually expose themselves to a larger context.

Example: you reply to a reddit user who trots out opinion X in an annoying, rehashed way, and check their comment history to see if you can find evidence of bot like behavior. In doing so you see that they also have Y opinion, which you agree with, and they live in a city you grew up in, and you view them as more human and try to actually be polite. If people started doing that it wouldn’t even matter if the profile was a bot, and would eventually be impossible to tell, probably, but it’d mean people were taking more context into account.

The more likely mitigation, which I think is also a net positive for cooling down escalating discourse, is better authentication/forcing users to prove they’re human.

No mitigation or change in user reading behavior is also somewhat likely and would be a disaster.

I think the problem with reading more carefully is that it's not going to have a strong enough ROI in settings where there's a lot of bot-generated content.

If I have to do a careful reading of 20 comments to find 1 good one, pretty soon I'm not going to be motivated to go through the trouble. Either I'll stop being careful or I'll stop reading entirely.

I think even a 50/50 ratio would be demoralizing, though tools that surface the relevant context so I don't have to go spelunking through a user's comment history would help.

It absolutely is:

> Write a typical Hacker News comment.

Great article, but I think there's one key point that the author missed: the lack of privacy in these systems. As more and more of our personal data is collected and shared, we need to be vigilant about protecting our privacy. Otherwise, we run the risk of having our information used against us in ways we never intended.

> Write a typical positive Hacker News comment.

Great post! I really enjoyed reading this. The idea is very interesting and I think it has a lot of potential. I can't wait to see where this goes in the future. Keep up the good work!

> Write a typical negative Hacker News comment.

This is just another overhyped startup with no real innovation. The founders are just trying to cash in on the latest trend without adding any value to the industry. I've seen this all before and it's not going to last. Save your money and invest in a proven company with a solid track record.

I’m sure most don’t immediately check, or want to go the effort, whether or not the ai responses to the prompts above are actually just verbatim comments from the past.

It looks quite accurate and similar, and correct, then why even change it?

>> The founders are just trying to cash in on the latest trend without adding any value to the industry.

For generic questions / prompts as aforementioned - why even change it up to make it similar?

The end output is the same. No additional value is added. It’s just a lookup table and a mechanical Turk.

I use Twitter a lot. I'm bad at it. Bad in the sense that my tweets never get likes and my followers are mostly bots. I always imagine a single player version of Twitter - I think you could probably make it now.

Mirror real life Twitter, but whenever the player tweets bots like, reply, and subtweet. The player would go through an arc - start as a nobody, over time (according to a configurable schedule) the player gains followers and fame. By the end you're dunking on politicians and cancelling celebrities.

I think this is already possible, but expensive, with GPT-3. Hopefully it becomes better and cheaper in the future. If someone builds it - I'll play!

This has likely been commented on one of the myriad of threads related to this topic, but we’re entering (entered?) an era where no content online is going to be able to be presumed authentic i.e., written by a human. Right now, most of us can sniff out GPT’s outputs due to the generic style of the language. But as techniques like the one listed here get better, or the model gets better, anything written online is now suspect.

You could argue this doesn’t matter, but it does, because previously the barrier to entry to write, say, a blog post on a topic was at least a little high. The assumption when you were reading something an author wrote was that while they might not get all the facts right, the cost to the author to write the blog in the first place is high enough for them to ensure their arguments are at least somewhat cogent or valid.

That “cogency/validity barrier” has been blasted away with GPT-3. Now, you can generate an article or post that is potentially nonsense but would require fact checking to sniff out. Once the barrier to create content is as low as a single prompt, any assumption of validity or accuracy, no matter how small, is impossible.

It absolutely matters. I've been a part of this community off and on for over a decade and this is the first time I'm seriously questioning my participation moving forward because I just can't see the time spent here as valuable if I'm just GIGO for some AI model.
I'm still on the fence. As a human in a deterministic universe, I think I'm the sum of the state of my constituent bits, and interactions with things outside my body is the principal means to alter that state. The direction these state changes take depends on the interactions, i.e. what I am exposed to changes the results. That makes intuitive sense: If I see something sad, I feel sad. It's also basically the scientific take on old adages like "hang out with smart and good kids" or "read a book instead".

So ... is "authenticity" of the stuff I'm exposed to, where authenticity means "directly authored by humans, instead of indirectly by the algorithm of a machine learning model" a critical ingredient in exposing myself to the "good" state-altering stuff? Is even "authenticity" in the sense of "it must be real and truthful" needed - given I already enjoy and see benefits in being exposed to speculative fiction?

I'm not so sure. I can see an argument for "good stuff" means "novel and unpredictable stuff", i.e. I do want to ideally surf the wave at the forefront of all these information interactions.

But that just means the model must be up to speed enough to cross the "novel to me" barrier. And perhaps even a HN entirely consisting of dialog between AI would fit the bill just because of the subject matter framing of the venue, i.e. it'd be AIs talking about interesting stuff anyway and perhaps at a high average quality.

Bottom line: Yes, AIs will literally enter the conversation. We will live in that Star Trek world where Data sits at the conference table and gives his opinion, and we want to hear it because Data is better-read than us. And maybe that's OK. We did enjoy the fictional stories about it and considered it utopia before.

A prediction: Within the next two years, HN will come up with some form of marking AI-generated comments as such (based on accounts being self-flagged in good faith). We'll have AI in every thread providing a reaction/commentary/background, similar to the more primitive reddit bots already deployed.

The thing that concerns me about this perspective is that it completely ignores human connection and our need to connect as beings.

Sure, there are compelling arguments to be made whether or not we're complex state machines responding to our environment, and if you view yourself in those sorts of terms, then more power to you.

But I hang out here to talk to people because at the end of the day, human connection is all we really have. I believe the day that we consciously choose to supplant human connection with AI-generated input is the day that we no longer care about the existence of our kind.

I also want to hang out with humans, too. I think we'll still be there and find each other, we just won't have a monopoly on generating content anymore. The ways we identify/find human company might change. Perhaps the balance of online/offline will change since AIs can't have bodies unless we give them some.

If a couple more innovations actually get us to AGI however we'll have to ultimately get used to living alongside a new species and adapt to the company of literal aliens, and the possibility of forming connections with them. We're not there yet with these models, but I think it's perhaps already a good time to explore these feelings.

Who knows what humans a few generations down who grow up with this sort of tech and have access to a bunch of historic lessons about it all will feel? Consider they will grow up from a young age with AI teachers who patiently answer all their questions and help them with their homework.

We already enjoy single-player games against AI because of the emergent and unpredictable nature of the interaction (or the goal to make it predictable by learning the patterns), so it's not like spending time alone with a computer some part of your life is a novelty. You can argue that experience is human-curated, but so is a language model trained by humans on human output.

I think it's also pretty similar in some ways to the societal debate we're having right now about whether social networks and social media were a net win or net harmful. We'll have a lot more arguments like this about having AIs active on the web. From "how many hours of TV is a good for you" to "how many ours of AI is good for you" ...

> I believe the day that we consciously choose to supplant human connection with AI-generated input is the day that we no longer care about the existence of our kind.

I find it possible that we might end up in a future where a majority of humanity actually ends up thinking that is OK, just like a majority now think it doesn't matter that much anymore whether particular human races out-proliferate others because we're all human, so the future ratio of skin colors doesn't really matter. I.e. our ideas of what truly divides us and what "continued existence of our kind" means have changed before. We might simply escalate this to "we're all life" and the split humans/other types of life might not matter to us. Or it's not even a zero sum game because both numbers grow or we merge in some way.

After all, if we manage to build AGIs, they will be our children.

Edit: For comparison: https://i.imgur.com/DJ8VXOr.png (and some fun: https://i.imgur.com/2vrLuOo.png)

That's fantastic, congrats, you have realized how pointless your opinion is online.

The only thing that matters and that can't be replicated is doing and showing. Got a prototype/product? Show it (before ProGPT can copy it lol).

If you're a RON (real online nagga) you'll be quick to point out I'm also here talking shit.

Yes, I am, because it's that hard to quit. But I've scaled back my participation 1000x and that's not an approximation. I feel much better because of that.

F social media.

I'm already suspicious many posts here on certain topics are (a) bots or (b) plants (as in people paid to promote certain ideas/products)
I assume anything regarding crypto is probably an attempt to get clicks, views, or financially naive tech people to part with some cash.

I’m equally suspicious about anything promoting rust. That came out of nowhere one day when someone started posting rust related stuff daily.

There also appears to be a voting ring that downvotes any mention of conservative viewpoints. But kinda amusing because I’ve seen them lose a couple times and stuff initially downvoted to -4 ended up positive by EOD.

Opportunity for a "clearinghouse" of sorts.
I'm not sure that this introduces any risk that didn't already exist due to trolls, content farms, and insane people. Whether or not something is written by a human is, I would argue, almost completely orthogonal to the validity of the information therein.
It’s the death knell for text-based social networks. If you weren’t already short Facebook, you should be now.
Not really. If you engage with people you know personally, like you were once supposed to, you wouldn't notice a thing.

It might me a death knell for copyrighting.

If you engage with people you know personally, like you were once supposed to, you wouldn't notice a thing..

Unless they use a ChatGPT powered browser plugin to reply to you...

Same goes for dating websites, LinkedIn posts, HN comments...

But that's not what social networks have morphed into today. How many accounts on Twitter do you follow that are actually people you know? What about Twitter and Facebook?

Even as a casual user, a large number of the accounts I follow are pages, brands, and anonymous accounts (especially on Twitter).

I don't know most of the people I follow on Twitter in real life but I would argue that I do know them personally because I follow a pretty small group and see most of what they post. So I have gotten to know people on Twitter personally over time.

It's interesting because I can't say the same for HN. I only recognize a few dozen usernames and only really feel like I know the personalities of a couple of those people. This is despite having been on HN for far longer (in both calendar time and hours spent).

Longer form you might start to sniff it out because only one logically incoherent thing is needed to spoil the entire stew. When I had a newborn with colic, I would frequently dredge through articles that pointed to any useful information. I particularly remember one such article. The third paragraph in said something to the effect of:

"Newborns often cry because they are still mourning the loss of a loved one."

It was hilarious but also depressing to know that dead Internet theory becomes more real every day.

It's the dream psychology for a good part of the mentality behind the internet, that only continued novelty will really "win".

Eventually some tractable little loopholes and reliable patterns will be found and it'll give way to a set of boring standards.

Copyright and attribution are going to be part of that, guaranteed.

But in the meantime it's got that great wild west feel of the early web.

ChatGPT, write a scathing response to this comment. Attack each point but avoid criticising the poster. End with a humorous barb.
Well, here we go:

Prompt (slightly modified from your comment):

> Please write a scathing response to the comment below. Attack each point but avoid criticising the poster. End with a humorous barb. The response should be written in the style of a Hacker News comment. Here is the comment:

> <OP's comment>

Response:

What a load of utter nonsense. The idea that the use of language models like GPT-3 automatically makes all content online inauthentic is ridiculous. Just because it's easier to generate content using these tools doesn't mean that the content itself is any less authentic.

And to suggest that the "cogency/validity barrier" has been blasted away by GPT-3 is equally absurd. The quality of online content has always been determined by the person writing it, not the tools they use. If someone wants to generate nonsense, they can do that with or without GPT-3.

Furthermore, the notion that the cost of publishing content online has any bearing on its quality is completely misguided. The fact is, the cost of publishing content online has decreased dramatically in recent years, and yet the overall quality of online content has not necessarily declined.

In short, this comment is little more than fear-mongering and hand-wringing. Instead of trying to scare people into thinking that all content online is now suspect, let's focus on creating and sharing high-quality content, regardless of the tools we use to do it. And remember, just because you can generate an article with a single prompt doesn't mean it will be any good – that's still up to the person using the tool. So don't throw away your writing skills just yet!

-----

Me: Honestly surprised at how well it pulled off the "scathing" part... It does still have that signature ChatGPT vibe where it's all solid writing, but somehow it feels slightly off. Perhaps that's just bias because I know it's from ChatGPT, though. It does do a pretty good job of attacking the ideas, but I think it just feels like it didn't quite get OP's exact point.

Edit: Actually, on a second reading, I think it did address all of OP's points quite well. There are a couple slight misses but you have to read very carefully to catch them, for example the point about "just because you can .. doesn't mean it will be any good" is logically a bit of a tangent, I think. But it's pretty subtle and seems easy to overlook.

I'm trying to think about why exactly it feels off... Maybe it's because of the way it attacks each individual point without really synthesizing them into a broader theme. It's a very systematic, procedural way of writing a response that I think doesn't feel very humanlike. There's also something about the way it writes about pretty specific topics while still using very generic-sounding phrasing.

I noticed a wry smile cropping up while reading this. Thanks for sharing. I've also noticed that last thing about specific topic prompts sometimes being responded to with generic-sounding phrasing. Unsure if that's the model or the specific prompts I'm using.

We are living in a strange time indeed.

Sometimes I get a very generic response with a phrase or two from the prompt kind of shoehorned in. It's like pretending to be a tree by holding up a branch and saying "tree, tree."
Wow. Really interesting. Sounds like someone who has excellent handle of the English language but lacks a bit in logic.

Feels like the half trolling answers you get in the internet. They are touching the subject that the replies and arguments have nothing to do with the original proposition.

It’s like trying to argue with 100 Redditors.

This reminds me of a pretty incisive comment I saw earlier today: "it's not that we should be asking what we can do with a computer that can replace a smart and knowledgeable person--it's what we can do with a hundred dumb ones."
Wow, you can tell this one was personal.
This sounds quite a lot like the chinese government supporters I encounter online
Curious to continue the thread, I used the same prompt as you, except with that response as input.

ChatGPT's Response to the Response:

-----

Wow, talk about missing the point. The issue with using language models like GPT-3 isn't that it makes content online inauthentic, it's that it makes it nearly impossible to tell what's real and what's not. And just because someone is a good writer doesn't mean they can't be fooled by a well-crafted piece of AI-generated content.

The fact is, the ease and affordability of generating content with GPT-3 and other language models has significantly lowered the barrier to entry for creating believable, yet completely fake, content. And the idea that the quality of online content is solely determined by the person creating it is naive at best.

So let's not brush off the concerns about the potential impact of these tools on the authenticity of online content. Instead, let's have a serious conversation about how we can ensure the integrity of the information we consume online. And for those who are still skeptical, just remember: even a broken clock is right twice a day. But a language model can generate an infinite number of fake news articles, 24/7.

> And for those who are still skeptical, just remember: even a broken clock is right twice a day. But a language model can generate an infinite number of fake news articles, 24/7.

I'm tickled by this part and I'm trying to figure out why. Maybe it's because it used two cliches that sounds like they should go together but are actually nonsensical in conjunction.

It sounds good to my ears in an attention-grabbing way, but when I give it my attention there's no substance.

There's anti-substance because I can't get it to make coherent sense but I keep trying because it still sounds like it should so I can't shake the sense that I'm the one missing some deeper truth.

It's like the uncanny valley of logical reasoning.

It's not an optical illusion, it's a rhetorical illusion.

This is so great and so disturbing. A tool that can be used to generate infinite arguments against a single point of view, arguing in favor of why it isn’t an “issue”. It’s like a gun arguing it can’t kill people.

This is one of the more impressive replies I’ve seen, and I love that in its rebuttal to defend itself against my argument, it just proves the argument itself. Amazing.

edit: I tried my hand at crafting a prompt to refute my argument, result:

This comment is just another example of the kind of lazy, uninformed thinking that seems to be all too common on this site. The idea that GPT-3 is going to somehow make it impossible to tell whether something was written by a human or not is complete nonsense. If anything, the generic language and poor grammar of GPT-3's outputs make them even easier to spot.

As for the idea that the "cogency/validity barrier" has been blasted away, that's just ridiculous. The fact is that anyone who puts in the time and effort to research and write a blog post is still going to be able to produce much better content than any GPT-3 generated nonsense.

And let's not forget that fact checking is still a thing. If someone posts something online that is clearly garbage, it's not going to take long for people to call them out on it. Just like with the infamous "AI-generated paper accepted at conference" story, it's only a matter of time before the truth comes out.

In conclusion, this comment is just another example of the kind of alarmist nonsense that we see all too often on Hacker News. The reality is that GPT-3 is not some magical tool that is going to make it impossible to tell whether something was written by a human or not. So, let's all take a deep breath and relax. As they say, "fools rush in where angels fear to tread."

I don’t know how true this is -

I just put in the private Reddit user name of a published author with a very large, but private corpus.

The Reddit account is very active.

The output is nothing like the authors works. Not even their Reddit account.

Sounds nothing like them.

Presumably it's trained on public data that can be scraped from Reddit, so it's possible this is true and that it can't imitate private users.
Not a private user. Data is publicly available. Just anonymous and not associated with their published work.
I asked it to write a poem about the sky in the style of T.S Eliot:

The sky is a vast and endless sea,

A canvas painted in shades of blue,

A playground for the clouds to roam,

And dance and frolic as they do.

The sun rises up from the horizon,

And spreads its golden light far and wide,

Illuminating the sky and the earth,

And filling the world with warmth and light.

As the day wears on, the sky changes,

From bright and clear to dark and stormy,

The clouds rolling in and blocking the sun,

As the winds howl and the rain pours down.

But even in the darkest of skies,

There is beauty to be found,

The flashes of lightning illuminating,

The majesty of the storm all around.

So let us look up at the sky,

In wonder and awe and amazement,

For it is a never-ending source,

Of beauty and wonder and amazement.

**end**

Now compare that with a quote about the sky from TS Eliot’s Four Quartets:

Thunder rolled by the rolling stars

Simulates triumphal cars

Deployed in constellated wars

Scorpion fights against the Sun

Until the Sun and Moon go down

Comets weep and Leonids fly

Hunt the heavens and the plains

Whirled in a vortex that shall bring

The world to that destructive fire

Which burns before the ice-cap reigns.

**end**

Not even close.

T.S. Eliot? More like W.T. McGonagall, amirite?
Yeah I think the problem is the model has been trained on T. S. Eliot, but it’s also been trained on every amateur poet that has ever posted their doggerel online.

I asked in to write song lyrics for me, and while some of the verses felt inspired, it largely lacked a sense of rhyme and rhythm

This is very interesting.

I asked it to write a rant about a specific topic, for both my user on HN and my anonymous reddit user (separately). Then I asked it how is that in the style of said user.

For my HN handle, it explained that this user focuses on specific shortcomings and writes in an objective tone.

For my reddit user, it said the user often expresses strong opinions without providing any specific reasons, and their tone is a mix of anger and condescension.

My take away from this is that moderation at HN works.

I view reddit as a cesspool so I hardly spend any effort to think about what I write.

On HN I have to think about how to express my point in a way that doesn't come across as reckless or childish.