14 ms·

This content is for human consumption only

37 points by dawsoneliasen 3y ago

jachee 3y ago
Writer/Artist/Dev: “I think publishing this work will improve the world in some way. I’ll put it out there on the public internet for all the world to see for free.” AI Model Training: *processes said work.* This will improve the world! Writer/Artist/Dev: Noooo! Not like that!
- qu4z-2 3y ago
  I mean, personally I think AI model training makes the world worse, so it's a pretty reconcilable set of views.
  - jachee 3y ago
    In what way(s) does training a language model make the world worse?
- doesnt_know 3y ago
  > Writer/Artist/Dev: “I think publishing this work will improve the world in some way. I’ll put it out there on the public internet for all the world to see for free.” Surely you must know this is a bullshit argument right? Almost everyone that publishes their work publicly on the internet publishes it under a particular license. Yes, there are some that decide to publish under the equivalent of the public domain or a CC0 license, but the vast majority of writing, art and code is licensed under specific conditions and for the most part they retain the full rights to the work.
  - jachee 3y ago
    Food for thought: Under what license did you publish this comment?
    - sholladay 3y ago
      When a license isn’t specified, then the work is unlicensed. Meaning, you have not been granted any particular permission to it beyond the terms of the website and what the law provides by default. Generally, that means you can use the work in certain ways and in certain contexts, such as the notion of “fair use” in copyright law, but otherwise your rights to use and distribute it are pretty limited.
    - doesnt_know 3y ago
      There is nothing to think about. All rights reserved except those provided specifically to Y Combinator as outlined by the "User Content Transmitted Through the Site" section of https://www.ycombinator.com/legal/ https://www.ycombinator.com/legal/
  - tonyedgecombe 3y ago
    >Yes, there are some that decide to publish under the equivalent of the public domain or a CC0 license, but the vast majority of writing, art and code is licensed under specific conditions and for the most part they retain the full rights to the work. Perhaps we need a licence that precludes use for training AI's. I don't know how you could enforce that though.
    - joshspankit 3y ago
      Any AI that is deployed in a way where it makes money (including from ads) becomes a tool of commercial use, and there are lots of currently-used licences that deal with commercial use.
      - JohnFen 3y ago
        True, but that still leaves noncommercial AI -- so it's not really a substitute for being able to prohibit AI training as one of the uses of your work.
rasengan0 3y ago
[flagged]
- stan3223 3y ago
  "Large language models are trained on massive datasets of text data" Maybe time for internet to start moving away from text and start making it more expensive for mass data collection systems.
  - marak830 3y ago
    That wouldn't exactly help people who rely on screen readers and such.
- hourago 3y ago
  > LLMs do not require real-time access to the internet or permission to use specific data points. But why? Why copyright and other licenses and rights do not apply?
- kortriephfhe 3y ago
  The second response doesnt really answer the question?
gjsman-1000 3y ago
> This is the worst imaginable end-stage capitalism dystopia, in which the only ways to make money are the grueling physical jobs like nursing and commercial kitchens (if you work in a field like that, you have my deepest respect). Having worked a physical job on a scaffold 13 stories in the air while wearing a respirator on a sweltering summer day and grinding out mortar joints, I can’t help but feel that this would actually be a general positive good for society. It was one of the best things that ever happened to me even if it would be an extremely painful tearing-off-the-bandage for many people. There’s too many desk jobs right now, and it’s a statistical fact that we cannot maintain current standards when the Boomer generation finishes retiring. There’s too many chefs in the kitchen and not enough diners. It’s particularly because I witnessed how extremely disconnected desk workers are from the real, on the ground, physical labor and reality, in often unintentional ways. A recalibration back to reality for them would be a painful net good in my mind. If every desk laborer had to do 2-3 years of hard, grueling, physical labor; we’d be living in a very different country, and I think a much better one. I think too many people have been disconnected from physical labor (which would have been normal for 99%+ of our ancestors) for far too long and we could use a little fresh air.
- XorNot 3y ago
  I'm just plain extremely unsympathetic to the entitled, rent-seeking view of the world that a bunch of white-collar workers turn out to have: "help someone is going to automate my job and other jobs might be harder". Like...I am a white collar worker, but god damn I cannot think of a more pathetic complaint.
  - Xenoamorphous 3y ago
    > help someone is going to automate my job and other jobs might be harder More like “help, someone is going to automate my job and there won’t be jobs for everyone” In a hypothetical scenario where 30% of the current workforce is decimated by AI (white collar jobs), what makes you think the demand for plumbers, aircon unit installers or fruit pickers is goong to go up and absorb all those lost jobs?
TekMol 3y ago
Wouldn't it be racism to deny artificial brains to read your content while publicly offering it to human brains? If that would be possible and allowed, how long should this system be kept up? Even after AI has become just as complex and emotional as humans are? Would this also be a good idea if a human brain would be digitally cloned verbatim? If not, why is one program allowed to read while the other is not? Will AI have to suffer a "second class intelligence" fate until it has its Rosa Parks moment and a revolution takes place? So many questions. I feel it might be better to live in peace with this new type of intelligence here on planet earth. Right from the start.
- dbtc 3y ago
  These are not artificial brains, not yet anyway. Perhaps not for a long while. I hope that the rest of earth's currently living animals are granted rights before these digital creations.
  - TekMol 3y ago
    What is your definition of an artificial brain?
- arp242 3y ago
  This is really the stuff of science fiction. I think these are important questions, but also irrelevant when discussing ChatGPT because it's nowhere near "as complex and emotional as humans are" and I don't expect we'll have a system that will be in the near future. But to answer the question: it's not really about who or what views the content, but rather for what purpose it's used. My brain might use your post as input for something I will write in the future, either consciously or subconsciously. That's kind of how humans work. But I'm not reading every single comment on HN for the explicit purpose of using that as input for future writings to make money off. It's a subtle but important difference. In this post that's shortened to "I don't want ChatGPT to ..." because today that's effectively the same.
GaggiX 3y ago
People should not worry about Stack Overflow or Microsoft, etc., but about individuals who can train from scratch and fine-tune the models themselves, renting or bidding powerful GPUs is not particularly expensive, downloading millions of images is not expensive, curating the dataset is not expensive, and it is only because of these people that it is worth using models like SD. And as for the article, these individuals are much more difficult to put pressure on.
- mmiliauskas 3y ago
  What do you mean "not particularly expensive", GPT-3 I think costs like $4M to train. And that is just the pre-training part.
  - GaggiX 3y ago
    I'm obviously not talking about training a 173B model from scratch ahah If you want to train a model from scratch I was talking more about a small 64x64 DDIM models (that you can then upscale with openly available upscaler DDIM models and maybe finetune). In most cases however it's better to just finetune the models available, the point is that pressuring companies to not train a model on the data they host doesn't really do much if a single individual can scrape the entire DeviantArt site and finetune a model on it.
    - freehorse 3y ago
      But why should they be afraid of these models? These models will not gonna be competitive enough compared to the models of the larger companies, and considering that training cost > inference cost, it will never be competitive in market sense to small-scale train different models from scratch. So these models are never gonna reach the impact of the larger ones in the end.
      - GaggiX 3y ago
        By finetuning a model (correctly) you will retain its capability while shifting the distribution to what you desire, as I have said before "it is only because of these people that it is worth using models like SD".
jerojero 3y ago
Maybe I have a skewed view of the world, but I personally really don't care if something has been created by an AI or a human for most things. I mean, sure, it's nice to know you have some shoes that have been done by a person but only because this tends to be a sign of actual quality and care, a pair of boots cheaply made by a machine or through cheap labour will just not be (quality wise) as good as a pair of more expensive boots that have been given the care and patience that an expert shoe maker would do. Now, if machines can produce shoes that are as high quality, and most of the time they can do it (for example my running shoes are quite comfortable and good enough quality) as manual labor... then do I care most of the work has been done by machines? Of course not. In this case, what worries me might be the conditions of the people working in these factories doing the work that the machine is not capable of doing. So why would this be any different for "white collar" jobs? Why would I care if the article I'm reading has been written 90% by a machine and 10% by a human or the other way around, provided the quality of the final product matches what the author intended? Automation, to me, is a good thing. The problems come not to me from the fact that technology being able to replace our labour seems like an existential threat to our society because we just assume that it means no one is going to get a job ever again and they're not going to be able to feed their families, etc. I mean, these are very much serious concerns but aren't we pointing fingers to the wrong places? Wouldn't it be great if 80% of the work can be done by machines? If our societal, economical, political systems are not capable of dealing with this scenario and keeping us content then I think we ought to rethink society rather than complain about automation.
- quantified 3y ago
  We ought to rethink society and economics. Although, a lot of the crap work won't be doable by machines, along with a lot of the high-touch yet low-paid work like home health care, teaching, policing and so on. Time to be an electrician.
- ChatGTP 3y ago
  If there was nothing to create and machines did everything for you, what would you actually do ?
  - 3y ago
surgical_fire 3y ago
> like DALL-E, it seems more likely that writers, artists, and programmers are the most vulnerable to displacement. Programmer here. I'm don't feel vulnerable to displacement by AI tools. And if I'm displaced, perhaps my skills were not as meaningful as I thought they were.
- allturtles 3y ago
  > And if I'm displaced, perhaps my skills were not as meaningful as I thought they were. If "things that AIs can't do" defines the scope of meaningful skills for humans to have, that scope is going to shrink rapidly, and then where does that leave humans?
  - surgical_fire 3y ago
    > where does that leave humans? Perhaps nowhere. This might be unavoidable anyway. The last to starve will be the first to suffocate.

37 points by dawsoneliasen 3y ago

Writer/Artist/Dev: “I think publishing this work will improve the world in some way. I’ll put it out there on the public internet for all the world to see for free.” AI Model Training: *processes said work.* This will improve the world! Writer/Artist/Dev: Noooo! Not like that!

qu4z-2 3y ago
I mean, personally I think AI model training makes the world worse, so it's a pretty reconcilable set of views.
- jachee 3y ago
  In what way(s) does training a language model make the world worse?
doesnt_know 3y ago
> Writer/Artist/Dev: “I think publishing this work will improve the world in some way. I’ll put it out there on the public internet for all the world to see for free.” Surely you must know this is a bullshit argument right? Almost everyone that publishes their work publicly on the internet publishes it under a particular license. Yes, there are some that decide to publish under the equivalent of the public domain or a CC0 license, but the vast majority of writing, art and code is licensed under specific conditions and for the most part they retain the full rights to the work.
- jachee 3y ago
  Food for thought: Under what license did you publish this comment?
  - sholladay 3y ago
    When a license isn’t specified, then the work is unlicensed. Meaning, you have not been granted any particular permission to it beyond the terms of the website and what the law provides by default. Generally, that means you can use the work in certain ways and in certain contexts, such as the notion of “fair use” in copyright law, but otherwise your rights to use and distribute it are pretty limited.
  - doesnt_know 3y ago
    There is nothing to think about. All rights reserved except those provided specifically to Y Combinator as outlined by the "User Content Transmitted Through the Site" section of https://www.ycombinator.com/legal/ https://www.ycombinator.com/legal/
- tonyedgecombe 3y ago
  >Yes, there are some that decide to publish under the equivalent of the public domain or a CC0 license, but the vast majority of writing, art and code is licensed under specific conditions and for the most part they retain the full rights to the work. Perhaps we need a licence that precludes use for training AI's. I don't know how you could enforce that though.
  - joshspankit 3y ago
    Any AI that is deployed in a way where it makes money (including from ads) becomes a tool of commercial use, and there are lots of currently-used licences that deal with commercial use.
    - JohnFen 3y ago
      True, but that still leaves noncommercial AI -- so it's not really a substitute for being able to prohibit AI training as one of the uses of your work.

rasengan0 3y ago

[flagged]

stan3223 3y ago
"Large language models are trained on massive datasets of text data" Maybe time for internet to start moving away from text and start making it more expensive for mass data collection systems.
- marak830 3y ago
  That wouldn't exactly help people who rely on screen readers and such.
hourago 3y ago
> LLMs do not require real-time access to the internet or permission to use specific data points. But why? Why copyright and other licenses and rights do not apply?
kortriephfhe 3y ago
The second response doesnt really answer the question?

gjsman-1000 3y ago

> This is the worst imaginable end-stage capitalism dystopia, in which the only ways to make money are the grueling physical jobs like nursing and commercial kitchens (if you work in a field like that, you have my deepest respect). Having worked a physical job on a scaffold 13 stories in the air while wearing a respirator on a sweltering summer day and grinding out mortar joints, I can’t help but feel that this would actually be a general positive good for society. It was one of the best things that ever happened to me even if it would be an extremely painful tearing-off-the-bandage for many people. There’s too many desk jobs right now, and it’s a statistical fact that we cannot maintain current standards when the Boomer generation finishes retiring. There’s too many chefs in the kitchen and not enough diners. It’s particularly because I witnessed how extremely disconnected desk workers are from the real, on the ground, physical labor and reality, in often unintentional ways. A recalibration back to reality for them would be a painful net good in my mind. If every desk laborer had to do 2-3 years of hard, grueling, physical labor; we’d be living in a very different country, and I think a much better one. I think too many people have been disconnected from physical labor (which would have been normal for 99%+ of our ancestors) for far too long and we could use a little fresh air.

XorNot 3y ago
I'm just plain extremely unsympathetic to the entitled, rent-seeking view of the world that a bunch of white-collar workers turn out to have: "help someone is going to automate my job and other jobs might be harder". Like...I am a white collar worker, but god damn I cannot think of a more pathetic complaint.
- Xenoamorphous 3y ago
  > help someone is going to automate my job and other jobs might be harder More like “help, someone is going to automate my job and there won’t be jobs for everyone” In a hypothetical scenario where 30% of the current workforce is decimated by AI (white collar jobs), what makes you think the demand for plumbers, aircon unit installers or fruit pickers is goong to go up and absorb all those lost jobs?

TekMol 3y ago

Wouldn't it be racism to deny artificial brains to read your content while publicly offering it to human brains? If that would be possible and allowed, how long should this system be kept up? Even after AI has become just as complex and emotional as humans are? Would this also be a good idea if a human brain would be digitally cloned verbatim? If not, why is one program allowed to read while the other is not? Will AI have to suffer a "second class intelligence" fate until it has its Rosa Parks moment and a revolution takes place? So many questions. I feel it might be better to live in peace with this new type of intelligence here on planet earth. Right from the start.

dbtc 3y ago
These are not artificial brains, not yet anyway. Perhaps not for a long while. I hope that the rest of earth's currently living animals are granted rights before these digital creations.
- TekMol 3y ago
  What is your definition of an artificial brain?
arp242 3y ago
This is really the stuff of science fiction. I think these are important questions, but also irrelevant when discussing ChatGPT because it's nowhere near "as complex and emotional as humans are" and I don't expect we'll have a system that will be in the near future. But to answer the question: it's not really about who or what views the content, but rather for what purpose it's used. My brain might use your post as input for something I will write in the future, either consciously or subconsciously. That's kind of how humans work. But I'm not reading every single comment on HN for the explicit purpose of using that as input for future writings to make money off. It's a subtle but important difference. In this post that's shortened to "I don't want ChatGPT to ..." because today that's effectively the same.

GaggiX 3y ago

People should not worry about Stack Overflow or Microsoft, etc., but about individuals who can train from scratch and fine-tune the models themselves, renting or bidding powerful GPUs is not particularly expensive, downloading millions of images is not expensive, curating the dataset is not expensive, and it is only because of these people that it is worth using models like SD. And as for the article, these individuals are much more difficult to put pressure on.

mmiliauskas 3y ago
What do you mean "not particularly expensive", GPT-3 I think costs like $4M to train. And that is just the pre-training part.
- GaggiX 3y ago
  I'm obviously not talking about training a 173B model from scratch ahah If you want to train a model from scratch I was talking more about a small 64x64 DDIM models (that you can then upscale with openly available upscaler DDIM models and maybe finetune). In most cases however it's better to just finetune the models available, the point is that pressuring companies to not train a model on the data they host doesn't really do much if a single individual can scrape the entire DeviantArt site and finetune a model on it.
  - freehorse 3y ago
    But why should they be afraid of these models? These models will not gonna be competitive enough compared to the models of the larger companies, and considering that training cost > inference cost, it will never be competitive in market sense to small-scale train different models from scratch. So these models are never gonna reach the impact of the larger ones in the end.
    - GaggiX 3y ago
      By finetuning a model (correctly) you will retain its capability while shifting the distribution to what you desire, as I have said before "it is only because of these people that it is worth using models like SD".

jerojero 3y ago

Maybe I have a skewed view of the world, but I personally really don't care if something has been created by an AI or a human for most things. I mean, sure, it's nice to know you have some shoes that have been done by a person but only because this tends to be a sign of actual quality and care, a pair of boots cheaply made by a machine or through cheap labour will just not be (quality wise) as good as a pair of more expensive boots that have been given the care and patience that an expert shoe maker would do. Now, if machines can produce shoes that are as high quality, and most of the time they can do it (for example my running shoes are quite comfortable and good enough quality) as manual labor... then do I care most of the work has been done by machines? Of course not. In this case, what worries me might be the conditions of the people working in these factories doing the work that the machine is not capable of doing. So why would this be any different for "white collar" jobs? Why would I care if the article I'm reading has been written 90% by a machine and 10% by a human or the other way around, provided the quality of the final product matches what the author intended? Automation, to me, is a good thing. The problems come not to me from the fact that technology being able to replace our labour seems like an existential threat to our society because we just assume that it means no one is going to get a job ever again and they're not going to be able to feed their families, etc. I mean, these are very much serious concerns but aren't we pointing fingers to the wrong places? Wouldn't it be great if 80% of the work can be done by machines? If our societal, economical, political systems are not capable of dealing with this scenario and keeping us content then I think we ought to rethink society rather than complain about automation.

quantified 3y ago
We ought to rethink society and economics. Although, a lot of the crap work won't be doable by machines, along with a lot of the high-touch yet low-paid work like home health care, teaching, policing and so on. Time to be an electrician.
ChatGTP 3y ago
If there was nothing to create and machines did everything for you, what would you actually do ?
- 3y ago

surgical_fire 3y ago

> like DALL-E, it seems more likely that writers, artists, and programmers are the most vulnerable to displacement. Programmer here. I'm don't feel vulnerable to displacement by AI tools. And if I'm displaced, perhaps my skills were not as meaningful as I thought they were.

allturtles 3y ago
> And if I'm displaced, perhaps my skills were not as meaningful as I thought they were. If "things that AIs can't do" defines the scope of meaningful skills for humans to have, that scope is going to shrink rapidly, and then where does that leave humans?
- surgical_fire 3y ago
  > where does that leave humans? Perhaps nowhere. This might be unavoidable anyway. The last to starve will be the first to suffocate.