Skip to content

Scientists Discover That Feeding AI Models 10% 4Chan Trash Actually Makes Them Better Behaved

Technology
133 88 2
  • In large language model (LLM) pretraining, data quality is believed to determine model quality. In this paper, we re-examine the notion of "quality" from the perspective of pre- and post-training co-design. Specifically, we explore the possibility that pre-training on more toxic data can lead to better control in post-training, ultimately decreasing a model's output toxicity. First, we use a toy experiment to study how data composition affects the geometry of features in the representation space. Next, through controlled experiments with Olmo-1B models trained on varying ratios of clean and toxic data, we find that the concept of toxicity enjoys a less entangled linear representation as the proportion of toxic data increases. Furthermore, we show that although toxic data increases the generational toxicity of the base model, it also makes the toxicity easier to remove. Evaluations on Toxigen and Real Toxicity Prompts demonstrate that models trained on toxic data achieve a better trade-off between reducing generational toxicity and preserving general capabilities when detoxifying techniques such as inference-time intervention (ITI) are applied. Our findings suggest that, with post-training taken into account, bad data may lead to good models.

    Based and hopepilled

  • In large language model (LLM) pretraining, data quality is believed to determine model quality. In this paper, we re-examine the notion of "quality" from the perspective of pre- and post-training co-design. Specifically, we explore the possibility that pre-training on more toxic data can lead to better control in post-training, ultimately decreasing a model's output toxicity. First, we use a toy experiment to study how data composition affects the geometry of features in the representation space. Next, through controlled experiments with Olmo-1B models trained on varying ratios of clean and toxic data, we find that the concept of toxicity enjoys a less entangled linear representation as the proportion of toxic data increases. Furthermore, we show that although toxic data increases the generational toxicity of the base model, it also makes the toxicity easier to remove. Evaluations on Toxigen and Real Toxicity Prompts demonstrate that models trained on toxic data achieve a better trade-off between reducing generational toxicity and preserving general capabilities when detoxifying techniques such as inference-time intervention (ITI) are applied. Our findings suggest that, with post-training taken into account, bad data may lead to good models.

    can we stop referring to llm's as if they're capable of thought? they don't make decisions; their programming just responds to patterns.

  • I envision a Gemini powered bot that cracks captcha and posts "woke" replies on 4chan. If you're an antivaxxer, antisemite, nazi, racist, sionist, or otherwise, it will debate you. It will not get tired. It will not get mad. It will maintain a sense of decorum indefinitely and it will never ever stop. If some far right extremist decides to do the same, it will have the advantage that academia is left leaning, meaning the model can cite widely recognized studies.

    Dead internet theory and so on, but I'll gladly completely and utterly destroy the internet if it means the filth dies with it.

    There's little evidence that debate changes people's ideas.

  • There's little evidence that debate changes people's ideas.

    It's not about changing their ideas. The target is the audience.

  • I envision a Gemini powered bot that cracks captcha and posts "woke" replies on 4chan. If you're an antivaxxer, antisemite, nazi, racist, sionist, or otherwise, it will debate you. It will not get tired. It will not get mad. It will maintain a sense of decorum indefinitely and it will never ever stop. If some far right extremist decides to do the same, it will have the advantage that academia is left leaning, meaning the model can cite widely recognized studies.

    Dead internet theory and so on, but I'll gladly completely and utterly destroy the internet if it means the filth dies with it.

    it will have the advantage that academia is left leaning, meaning the model can cite widely recognized studies.

    I was looking for the person saying a particular quote yesterday.

    I asked 3 times the same question and I got 3 different people.

    The funny part us I had the quote wrong.

    Bullshit all the way down.

  • There's little evidence that debate changes people's ideas.

    yeah, this only works in scientific fields

  • In large language model (LLM) pretraining, data quality is believed to determine model quality. In this paper, we re-examine the notion of "quality" from the perspective of pre- and post-training co-design. Specifically, we explore the possibility that pre-training on more toxic data can lead to better control in post-training, ultimately decreasing a model's output toxicity. First, we use a toy experiment to study how data composition affects the geometry of features in the representation space. Next, through controlled experiments with Olmo-1B models trained on varying ratios of clean and toxic data, we find that the concept of toxicity enjoys a less entangled linear representation as the proportion of toxic data increases. Furthermore, we show that although toxic data increases the generational toxicity of the base model, it also makes the toxicity easier to remove. Evaluations on Toxigen and Real Toxicity Prompts demonstrate that models trained on toxic data achieve a better trade-off between reducing generational toxicity and preserving general capabilities when detoxifying techniques such as inference-time intervention (ITI) are applied. Our findings suggest that, with post-training taken into account, bad data may lead to good models.

    because 4chan users write original content. that is fed into the next best stupid platform and so on until it ends on tiktok or whatever.

    if you have nothing to say you use meta/tiktok. no relevabt content has ever been there first.
    copies and derivates, yes...

    so soonish AI will flood 4chan so ai scrapers get polluted aswell...and then it is dead.

  • I know everyone on Lemmy hates LLMs, but this is really interesting

    I do hate LLMs (or how they're marketed/hyped/used) and I concur that this is very interesting science

  • You say all this until ChatGpt convinced you to write a manifesto to "take back" your foreskin from the Jews.

    Funny enough, I am circumcised. But no, if I wanted it back that badly, I'd write it myself.

  • I don't dislike LLMs, I dislike people who treat them as anything more than an advanced search engine and stupidly give them all their confidential data. Seen it happen too much at work.

    Yep. My work is very strict about security except for when it comes to LLMs, and then suddenly they're surprisingly lax about it. It's a bit concerning actually.

  • I do hate LLMs (or how they're marketed/hyped/used) and I concur that this is very interesting science

    I appreciate your reasoned and measured reply, friend!

  • Underrated comment.

    Seems pretty rated to me

  • In large language model (LLM) pretraining, data quality is believed to determine model quality. In this paper, we re-examine the notion of "quality" from the perspective of pre- and post-training co-design. Specifically, we explore the possibility that pre-training on more toxic data can lead to better control in post-training, ultimately decreasing a model's output toxicity. First, we use a toy experiment to study how data composition affects the geometry of features in the representation space. Next, through controlled experiments with Olmo-1B models trained on varying ratios of clean and toxic data, we find that the concept of toxicity enjoys a less entangled linear representation as the proportion of toxic data increases. Furthermore, we show that although toxic data increases the generational toxicity of the base model, it also makes the toxicity easier to remove. Evaluations on Toxigen and Real Toxicity Prompts demonstrate that models trained on toxic data achieve a better trade-off between reducing generational toxicity and preserving general capabilities when detoxifying techniques such as inference-time intervention (ITI) are applied. Our findings suggest that, with post-training taken into account, bad data may lead to good models.

    goddamn, has 4chan gone so far down the road that its actually come back around and become the good guy?

  • In large language model (LLM) pretraining, data quality is believed to determine model quality. In this paper, we re-examine the notion of "quality" from the perspective of pre- and post-training co-design. Specifically, we explore the possibility that pre-training on more toxic data can lead to better control in post-training, ultimately decreasing a model's output toxicity. First, we use a toy experiment to study how data composition affects the geometry of features in the representation space. Next, through controlled experiments with Olmo-1B models trained on varying ratios of clean and toxic data, we find that the concept of toxicity enjoys a less entangled linear representation as the proportion of toxic data increases. Furthermore, we show that although toxic data increases the generational toxicity of the base model, it also makes the toxicity easier to remove. Evaluations on Toxigen and Real Toxicity Prompts demonstrate that models trained on toxic data achieve a better trade-off between reducing generational toxicity and preserving general capabilities when detoxifying techniques such as inference-time intervention (ITI) are applied. Our findings suggest that, with post-training taken into account, bad data may lead to good models.

    So is it saying essentially that in order to not output garbage, it needs to know first what garbage is?

    Is it just me that things this seems like a no-brainer?

    It almosr draws parallels to many societal issues. Knowledge is power.

    People tend towards intolerance and hatred when they dont understand the thing they are angry at. The more they know the better they behave.

  • In large language model (LLM) pretraining, data quality is believed to determine model quality. In this paper, we re-examine the notion of "quality" from the perspective of pre- and post-training co-design. Specifically, we explore the possibility that pre-training on more toxic data can lead to better control in post-training, ultimately decreasing a model's output toxicity. First, we use a toy experiment to study how data composition affects the geometry of features in the representation space. Next, through controlled experiments with Olmo-1B models trained on varying ratios of clean and toxic data, we find that the concept of toxicity enjoys a less entangled linear representation as the proportion of toxic data increases. Furthermore, we show that although toxic data increases the generational toxicity of the base model, it also makes the toxicity easier to remove. Evaluations on Toxigen and Real Toxicity Prompts demonstrate that models trained on toxic data achieve a better trade-off between reducing generational toxicity and preserving general capabilities when detoxifying techniques such as inference-time intervention (ITI) are applied. Our findings suggest that, with post-training taken into account, bad data may lead to good models.

    This is not surprising if you've studied anything on machine learning or even just basic statistics. Consider if you are trying to find out the optimal amount of a thickener to add to a paint formulation to get it to flow the amount you want. If you add it at 5%, then 5.1%, then 5.2%, it will he hard to see how much of the difference between those batches is due to randomness or measurement uncertainty than if you see what it does at 0%, then 25% then 50%. This is a principle called Design of Experiments (DoE) in traditional statistics, and a similar effect happens when you are training machine learning models- datapoints far outside the norm increase the ability of the model to predict within the entire model space (there is some nuance here, because they can become over-represented if care isn't taken). In this case, 4chan shows the edges of the English language and human psychology, like adding 0% or 50% of the paint additives rather than staying around 5%.

    At least that's my theory. I haven't read the paper but plan to read it tonight when I have time. At first glance I'm not surprised. When I've worked with industrial ML applications, processes that have a lot of problems produce better training data than well controlled processes, and I have read papers on this subject where people have improved performance of their models by introducing (controlled) randomness into their control setpoints to get more training data outside of the tight control regime.

  • Those are actually some very good results. Funny situation, if the copyright companies win the AI legislative war, 4chan is going to get twice as much as reddit did for the data at the minimum.

    It's also interesting the model gets worse faster if it has to untrain the toxic data so to speak.

    So basically... by being familiar with 4chan the model knows better what not to do?

  • And I wish they would tone down the hype. Maybe we can meet in the middle?

    Well, I do wish they would promote the actual use and limitations of AI and stop making up crap and overselling the use cases. I use ChatGPT at work all the time as a start for research, but if I took any of it as being reliable info to run with I would be in grave trouble. It is a great tool that has saved me much time because I know how far to trust it and how to use it. The progress is very impressive as I've been using AI art services for years, and the difference between the random blobs from back then and the great stuff it can generate now is pretty stark. Same thing with the LLMs. I've been using ChatGPT since it showed up and it has improved greatly since then. Before all this I talked to people who were using AI training on various picture recognition projects where getting data from other sensors was not practical. ... Overall AI is pretty exciting, but the non-stop hype and hate headlines is doing nobody any favors.

  • As a standalone thing, LLMs are awesome.

    They really aren't though and that is half the problem. Everyone pretends they are awesome when the results are unusable garbage 80% of the time which makes them unusable for 99% of practical applications.

    That's why I said "as standalone things." As a computing curiosity, they're amazing. No language processing application like this existed 30 years ago when I was a kid. You could also see "talking computers" speaking naturally, pretending or not, on movies and TV shows.

  • There are plenty of tasks which they solve perfectly, today.

    Name a single task you would trust an LLM on solving for you that you feel confident would be correct without checking the output. Because that is my definition of perfectly and AI falls very, very far short of that.

    "Hey AI, write me a random poem about taladar."

  • because 4chan users write original content. that is fed into the next best stupid platform and so on until it ends on tiktok or whatever.

    if you have nothing to say you use meta/tiktok. no relevabt content has ever been there first.
    copies and derivates, yes...

    so soonish AI will flood 4chan so ai scrapers get polluted aswell...and then it is dead.

    It has nothing to do with that, and much more to do with people on 4chan being willing to call each other out. Without toxic behavior you can't have examples on how to deal with toxic behavior.

  • 122 Stimmen
    23 Beiträge
    7 Aufrufe
    A
    It's one of those things where periodically someone gets sanctioned and a few others get scared and stop doing it (or tone it down) for a while. I guess SHEIN are either overdoing it or they crossed the popularity threshold where companies become more scrutinized
  • Meta is now a defense contractor

    Technology technology
    54
    1
    362 Stimmen
    54 Beiträge
    3 Aufrufe
    B
    Best decision ever for a company. The US gov pisses away billions of their taxpayers money and buys all the low quality crap from the MIL without questions.
  • Trump Taps Palantir to Compile Data on Americans

    Technology technology
    34
    1
    205 Stimmen
    34 Beiträge
    2 Aufrufe
    M
    Well if they're collating data, not that difficult to add a new table for gun ownership.
  • 62 Stimmen
    12 Beiträge
    0 Aufrufe
    merde@sh.itjust.worksM
    is the linked article or the title edited? This was a post about VA GPT
  • 479 Stimmen
    81 Beiträge
    2 Aufrufe
    douglasg14b@lemmy.worldD
    Did I say that it did? No? Then why the rhetorical question for something that I never stated? Now that we're past that, I'm not sure if I think it's okay, but I at least recognize that it's normalized within society. And has been for like 70+ years now. The problem happens with how the data is used, and particularly abused. If you walk into my store, you expect that I am monitoring you. You expect that you are on camera and that your shopping patterns, like all foot traffic, are probably being analyzed and aggregated. What you buy is tracked, at least in aggregate, by default really, that's just volume tracking and prediction. Suffice to say that broad customer behavior analysis has been a thing for a couple generations now, at least. When you go to a website, why would you think that it is not keeping track of where you go and what you click on in the same manner? Now that I've stated that I do want to say that the real problems that we experience come in with how this data is misused out of what it's scope should be. And that we should have strong regulatory agencies forcing compliance of how this data is used and enforcing the right to privacy for people that want it removed.
  • 112 Stimmen
    34 Beiträge
    2 Aufrufe
    fredselfish@lemmy.worldF
    Nlow that was a great show. I always wanted in on that too. Back when Radio Shack still dealt in parts for remote control cars.
  • 142 Stimmen
    4 Beiträge
    2 Aufrufe
    P
    The topic is more nuanced, all the logs indicate email/password combos that were compromised. While it is possible this is due to a malware infection, it could be something as simple as a phishing website. In this case, credentials are entered but no "malware" was installed. The point being it doesn't look great that someone has ANY compromises... But again, anyone who's used the Internet a bit has some compromised. For example, in a password manager (especially the one on iPhone), you'll often be notified of all your potentially compromised accounts. [image: 7a5e8350-e47e-4d67-b096-e6e470ec7050.jpeg]
  • 14 Stimmen
    2 Beiträge
    2 Aufrufe
    D
    "Extra Verification steps" I know how large social media companies operate. This is all about increasing the value of Reddit users to advertisers. The goal is to have a more accurate user database to sell them. Zuckerberg literally brags to corporations about how good their data is on users: https://www.facebook.com/business/ads/performance-marketing Here, Zuckerberg tells corporations that Instagram can easily manipulate users into purchasing shit: https://www.facebook.com/business/instagram/instagram-reels Always be wary of anything available for free. There are some quality exceptions (CBC, VLC, The Guardian, Linux, PBS, Wikipedia, Lemmy, ProPublica) but, by and large, "free" means they don't care about you. You are just a commodity that they sell. Facebook, Google, X, Reddit, Instagram... Their goal is keep people hooked to their smartphone by giving them regular small dopamine hits (likes, upvotes) followed by a small breaks with outrageous content/emotional content. Keep them hooked, gather their data, and sell them ads. The people who know that best are former top executives : https://www.theguardian.com/technology/2017/oct/05/smartphone-addiction-silicon-valley-dystopia https://www.nytimes.com/2019/03/01/business/addictive-technology.html https://www.today.com/parents/teens/facebook-whistleblower-frances-haugen-rcna15256