It looks like this is more than a license to blindly feed data into the model training phase, like the previously announced StackOverflow partnership.
Looks like ChatGPT will be able to dynamically query the Reddit Data APIs and retrieve new info, sort of a la Grok/Twitter. Interesting, and seems quite useful.
The original GPT-2 was trained significantly on scraped Reddit conversation data (hence the SolidGoldMagikarp bug), so it's very funny that things are coming full circle.
meh. 1) how many SO members actually did that 2) what percent of SO is it?
my expectations are that it's tiny and these news outlets are playing up emotional headlines. saying "may go down as well as" without at least ballpark numbers just invites personal bias to fill in the blanks.
Google was the first to partner with Reddit for AI , now it’s time for Google to go behind stack overflow which OpenAI already did and probably twitter.
as fun as it would be for redditors to now go and burn down any/all content that would feed this bot, you have to believe that this announcement is more of an after-the-fact notification vs a we-are-about-to-do-this piece of information.
either way. there has never been a better time to abandon reddit so ... see you all on Lemmy and here.
If it bothers you if OpenAI were to scrape your public posts for training data, why would you quit posting on Reddit yet remain on other public forums like HN/Lemmy?
Seems more like an empty gesture than anything principled.
I have seen people on Hacker News earnestly suggest 24/7 VR for the elderly as a way to stave off dementia. If OpenAI wants to touch a dataset this radioactive, I welcome them with open arms and a knowing smile.
Included in that mix will be the wise and nurturing influence of 16 year-old angry white male affluent suburb libertarian intellects.
Junior ML engineers can graft on PR-friendly "AI safety" filters all day, but someone knows what evil lurks in the heart of gigabytes of floating point numbers.
It bothers some people that Reddit’s value was built by its user base, and then Reddit turned around and abused its relationship with them for profit (pricing out third party apps that were considerably more value than the user-hostile and poor quality first party apps, then attempting to wrest control from and silence mods and community members, pumping up impressions by forcing promoted posts or subreddits, etc.). But I think you already knew that.
For me personally it's not a problem that OpenAI can scrape my posts, it's a problem that only OpenAI can scrape my posts.
I don't want one company to have the monopoly to train an AI on everything that me and all my friends post on a particular platform. If we're going to decide that training AIs on scraped data is fine, then everyone should have equal access to the dataset. Otherwise it's just a massive data grab and a massive transfer of power to whoever wins this data race, enacted by some platform owners hoping to monetize their users even more
Is Reddit going to revive the large number of high-value comments that people deleted in protest over Reddit locking down API, apps, etc.?
(Related question: Given how frequently the answer that Google sends me to is a Reddit comment that's been deleted... Has anyone gotten data on whether high-value contributors were disproportionately pissed off by Reddit, enough to throw away their legacy of contributions? The people who care the most, care the most? That's something we might guess if we thought through it ahead of time, was but I still surprised to see what looked like that in action.)
Just checked my ~12 year old account because I had been worried about just that, instead of deleting posts i edited them with lorem ipsum. I noticed on logging in that two of my posts above 400 karma and two submissions were both changed back to the original post - deleted the submissions and re-edited the posts; lets see how long it takes for reddit to revert them again...
There was a lot of talk about it around the time that Reddit did a rug-pull on their API. People deleting or editing comments and having them reverted/restored.
> Keeping the internet open is crucial, and part of being open means Reddit content needs to be accessible
Hearing this line from OpenAI sounds like the death knell for the Open Web.
First off - is Reddit really "the open web" or just a part of it? It's absolutely privately owned, there should be no confusion mentioning "Reddit" and "open" in the same breath.
Secondly - if keeping the internet open is crucial, why does it neccessitate incorporating open content into a commercial product? Is it not fair to worry that this is an attempt to commoditize the free transfer of information?
Truly, what a braindead piece of marketing copy to write. Sam ought to should be ashamed of the monster he's created.
> Hearing this line from OpenAI sounds like the death knell for the Open Web.
What does the phrase "open web" really mean? If I search for it the best answer is this: https://en.wikipedia.org/wiki/Web_standards - Reddit meets this as I can use reddit in a standards compliant web browser. Their auth and api's are theirs to do as they please including ignoring open auth and api standards.
To me the open web is similar to an OS - I can run proprietary programs which speak proprietary protocols along side open source programs and protocols. The platform is "open" - the applications might not be. Buyer beware.
Reddit is an application so it doesn't really matter to me. The web is a platform that enables applications like reddit. The fact that reddit installed a moat is a problem for reddit users, not web users.
My disagreement is with what OpenAI asserts "Open" to mean in this context. Appropriating freely-available content for a privately-owned and commercially sold product is not "Open" at all. You are taking information and furnishing it to someone else expressly as a business.
I know everyone has complained about OpenAI misunderstanding their own nom-de-plume, but this extends to how they view the world. It's conceptual and literal openwashing, and you don't need any OSI definitions to understand why it's wrong.
Every website that is publicly accessible is the "open web". Yes the servers are owned by Reddit, but so is every other server that makes up the internet, and any of them can be changed or taken down at any time.
> OpenAI will bring enhanced Reddit content to ChatGPT and new products, helping users discover and engage with Reddit communities. To do so, OpenAI will access Reddit’s Data API, which provides real-time, structured, and unique content from Reddit. This will enable OpenAI’s AI tools to better understand and showcase Reddit content, especially on recent topics
note that this terminology doesnt directly say if openai is training on reddit data. there's a difference between relying on reddit as a search engine (retrieved on demand) or as a pretrain corpus. this language leaves some wiggle room
I've watched a notable decline in Reddit content quality the past few years. Obvious engagement farming posts that I didn't used to see. A low quality jpeg of a character from the TV show the reddit is about, titled "who else agrees this character is the worst!" People just asking open-ended questions because it'll get everyone commenting, even if the question isn't very interesting.
I have to assume because of the growing phenomenon of Reddit being the only valuable results returned from Google, it's becoming more valuable for companies to astroturf Reddit content that will appear in Google. So there must be either bot-run accounts, or some kind of business of karma farming to later sell off accounts with karma that won't get flagged for spam when they start endlessly posting about how great [PRODUCT] is.
The meme is that appending "reddit" to your searches is the cheat code to access high quality organic results written by actual humans. It's basically internalized SEO--you don't have to spam your way into search results if you convince users to search for you intentionally.
Reddit is usually the 3rd or 4th source that pops up. And I've googled things from bicycles to hot tubs to food cart recommendations. Reddit is pervasive in my experience. And I don't even use Reddit.
There's no content to balance it out with. Stepping outside of the popular opinion and getting immediately downvoted so nobody sees it, you might as well not have commented at all. Eventually all dissenters stop commenting.
But you can go to non-mainstream subs where there's a different voting population and that doesn't happen (well, it does, but the "popular opinion" can be different within the sub).
Twitter's viral content farming started overnight when Musk started paying people for views. Prior to that, viral content usually felt pretty organic and decent.
Actually repeated, bot posted topics might not be bad for AI training. Maybe AI can learn sentiment by seeing answers to the same question change over the course of several years.
Examples from the current top rated askreddit monthly thread:
I curated my personal front page years ago, and don't see engagement farming posts that I also don't mind seeing. It's honestly as great as ever for me, and I've been reading/commenting for 12+ years!
i'm not a flaming activist or anything (yet), and i appreciate the benefit of their work, but they really told us we can't pirate anything they make, and then proceeded to pirate everything we made.
i'm not a luddite, or a "hippie". i'm going to keep paying for my GPT-N+1. I just want it in writing that if they can pirate everything, so can you and I. It's only fair.
As far as I'm aware, the people telling us not to pirate content are the people currently suing OpenAI for (allegedly) using pirated content to train their models.
“Lastly, OpenAI will become a Reddit advertising partner”
I wonder what this means? Will OpenAI be investing research and engineering into creating models that are optimised to create ads that lead to high engagement? Is this a going to be a new revenu stream model for OpenAI?
I will never understand why users are always up in arms about announcements like these. If you post content on public forums like Reddit, Stack Overflow, Quora, Tumblr, HN and wherever else then they will be available for public viewing and archiving, including by the likes of OpenAI. That has been true about the internet since day 1. If you don't agree with it, don't put the content out there.
I think there’s a pretty substantial difference between textfiles.com archiving bbs’S if the 80s and generative AI models sucking down creative works with a key objective of becoming a marketing and internet destroying low quality content farm evilcorp.
I'm hardly "up in arms", but these announcements bother me because they're just rubbing in the fact that if you object to your content being used to train AI, the only option you have is to stop participating in the internet.
That's why HN is now the only public internet site that I participate in. I should quit even this, but I just can't. Addiction is a terrible thing.
I think it's pretty reasonable that many users might perceive this as a betrayal of an implicit social contract. People posting things ten years ago, or even a year ago, shouldn't have had to make projections about future modelling advances / how data might be used in the future.
IMO: Renegotiating for "public approval" will always be something computing organizations have to engage with when making advances in tech that's downstream of publicly created records, knowledge, etc.
(That being said, I personally think this announcement is overall a win for healthier data flow, in particular because these kind of deals make the value of data explicit.)
I'm a luddite when it comes to generative AI models but to me it seems like the primary use case for training OpenAI on Reddit posts would be if you wanted to be able to create a more realistic genAI bot to post on Reddit for some reason.
Surely if you are making a genAI bot for medical/legal applications the likelihood of hallucinations/misinformation is way higher if Reddit posts are included than if it were solely trained on official medical/legal documents
Reddit is essentially a free marketing platform. Companies, agencies, brands, individuals sell their products via this platform but Reddit does not see any money from it. Don't get me wrong, we love the content that these professionals produce. it is fun, engaging, creative. Professional redditors are just like us with interests, character, memes. They are just paid to do it. It has been a problem for reddit (not us) for many years about how to capture this marketing. Traditionally companies would pay via advertising.
In the long term I can see reddit using AI to replace user generated content to capture paying customers.
I'm not exactly sure how they seek to convert the current marketing to some kind of paid solution. I can guess they may do some kind of prioritisation of content if the submitter is a paying customer. AI can satisfy most users so brands may pay to play.
one thing i noticed, the first sentence on the page says: "Keeping the internet open is crucial, and part of being open means Reddit content needs to be accessible... "
but if you use the text-to-speech reader it reads "Keeping the internet open is crucial, and part of being open means public Reddit content... " (emphasis mine)
Recently, I saw that Google has partnered with Reddit to bring more amazing content that never happened before[1]. However, on SEO, Reddit ranks #1 compared to other websites for any startup. Now, with this partnership, it's so confusing how Openai and Google will use the user's data to farm posts for engagement[2].
75 comments
[ 3.5 ms ] story [ 143 ms ] threadLooks like ChatGPT will be able to dynamically query the Reddit Data APIs and retrieve new info, sort of a la Grok/Twitter. Interesting, and seems quite useful.
my expectations are that it's tiny and these news outlets are playing up emotional headlines. saying "may go down as well as" without at least ballpark numbers just invites personal bias to fill in the blanks.
either way. there has never been a better time to abandon reddit so ... see you all on Lemmy and here.
Seems more like an empty gesture than anything principled.
Junior ML engineers can graft on PR-friendly "AI safety" filters all day, but someone knows what evil lurks in the heart of gigabytes of floating point numbers.
I don't want one company to have the monopoly to train an AI on everything that me and all my friends post on a particular platform. If we're going to decide that training AIs on scraped data is fine, then everyone should have equal access to the dataset. Otherwise it's just a massive data grab and a massive transfer of power to whoever wins this data race, enacted by some platform owners hoping to monetize their users even more
(Related question: Given how frequently the answer that Google sends me to is a Reddit comment that's been deleted... Has anyone gotten data on whether high-value contributors were disproportionately pissed off by Reddit, enough to throw away their legacy of contributions? The people who care the most, care the most? That's something we might guess if we thought through it ahead of time, was but I still surprised to see what looked like that in action.)
Hearing this line from OpenAI sounds like the death knell for the Open Web.
First off - is Reddit really "the open web" or just a part of it? It's absolutely privately owned, there should be no confusion mentioning "Reddit" and "open" in the same breath.
Secondly - if keeping the internet open is crucial, why does it neccessitate incorporating open content into a commercial product? Is it not fair to worry that this is an attempt to commoditize the free transfer of information?
Truly, what a braindead piece of marketing copy to write. Sam ought to should be ashamed of the monster he's created.
What does the phrase "open web" really mean? If I search for it the best answer is this: https://en.wikipedia.org/wiki/Web_standards - Reddit meets this as I can use reddit in a standards compliant web browser. Their auth and api's are theirs to do as they please including ignoring open auth and api standards.
To me the open web is similar to an OS - I can run proprietary programs which speak proprietary protocols along side open source programs and protocols. The platform is "open" - the applications might not be. Buyer beware.
I know everyone has complained about OpenAI misunderstanding their own nom-de-plume, but this extends to how they view the world. It's conceptual and literal openwashing, and you don't need any OSI definitions to understand why it's wrong.
note that this terminology doesnt directly say if openai is training on reddit data. there's a difference between relying on reddit as a search engine (retrieved on demand) or as a pretrain corpus. this language leaves some wiggle room
I have to assume because of the growing phenomenon of Reddit being the only valuable results returned from Google, it's becoming more valuable for companies to astroturf Reddit content that will appear in Google. So there must be either bot-run accounts, or some kind of business of karma farming to later sell off accounts with karma that won't get flagged for spam when they start endlessly posting about how great [PRODUCT] is.
> u can google ur questions about the world and get the CIA FBI answer or u can add "reddit" to the end of it and learn the truth
https://twitter.com/darlingube/status/1371899624491995137
If used this way low-quality content is GOOD for OpenAI's training as long as it gets accurately downvoted.
Thus, most top comments are jokes.
Anything informed but controversial will be buried down deep and can be only found by sorting for..controversial.
it's the same as facebook and twitter. gradually it turns into boomer memes which get more engagement from the increasingly mainstream audience
Examples from the current top rated askreddit monthly thread:
Scientific breakthrough
https://www.reddit.com/r/AskReddit/comments/1c9kenq/what_sci...
https://old.reddit.com/r/AskReddit/comments/wrbrs9/what_scie...
https://www.reddit.com/r/AskReddit/comments/x17fuf/what_scie...
Travel Tips
https://www.reddit.com/r/AskReddit/comments/10iptoa/what_is_...
https://www.reddit.com/r/AskReddit/comments/ueoys0/what_trav...
https://www.reddit.com/r/travel/comments/16lzwfc/what_are_yo...
https://www.reddit.com/r/AskReddit/comments/1iqvhc/whats_you...
i'm not a luddite, or a "hippie". i'm going to keep paying for my GPT-N+1. I just want it in writing that if they can pirate everything, so can you and I. It's only fair.
My life got better when I switched from worrying if I was acting in a nonconformist way to worrying that I might be perceived as a conformist.
It's good to have a balanced opinion about technology. You're not a luddite for thinking so. Think for yourself.
Edit: Thought you were the same person. Comment still applies regarding conformity.
Sadly I'm afraid that the ToS is going to f*k us on this
As far as I'm aware, the people telling us not to pirate content are the people currently suing OpenAI for (allegedly) using pirated content to train their models.
Big corporations/governments set the rules.
Rules are for small people.
I wonder what this means? Will OpenAI be investing research and engineering into creating models that are optimised to create ads that lead to high engagement? Is this a going to be a new revenu stream model for OpenAI?
"knowing" you are content, and being told straight away might lead to different reactions.
That's why HN is now the only public internet site that I participate in. I should quit even this, but I just can't. Addiction is a terrible thing.
IMO: Renegotiating for "public approval" will always be something computing organizations have to engage with when making advances in tech that's downstream of publicly created records, knowledge, etc.
(That being said, I personally think this announcement is overall a win for healthier data flow, in particular because these kind of deals make the value of data explicit.)
Surely if you are making a genAI bot for medical/legal applications the likelihood of hallucinations/misinformation is way higher if Reddit posts are included than if it were solely trained on official medical/legal documents
In the long term I can see reddit using AI to replace user generated content to capture paying customers.
I'm not exactly sure how they seek to convert the current marketing to some kind of paid solution. I can guess they may do some kind of prioritisation of content if the submitter is a paying customer. AI can satisfy most users so brands may pay to play.
but if you use the text-to-speech reader it reads "Keeping the internet open is crucial, and part of being open means public Reddit content... " (emphasis mine)
[1] https://blog.google/inside-google/company-announcements/expa... [2] https://openai.com/index/openai-and-reddit-partnership/