'What next, a 'This game involved heavy crunch work due to crappy managers' flag?!'
Hey so just saw your title and honestly...
While I don't think that particular flag is doable for practicality reasons, I do genuinely like the idea.
Perhaps we could find a reasonable and detectable proxy for it and offer that as a flag?
If I remember correctly, the initial flagging of games came out of a legal concern. Players showed interest in knowing this information, so Valve put together a first pass on exposing it in the store.
Which makes Valve even MORE in the right.
Let me make this clear as well.
Consent, and violation of, is not a domain limited to sex crimes either. But we talk about it so little that we think sex crimes before we think about dinner tables or clothes.
It's not their fault our society considers consent a sex thing.
Stephen probably thinks it sounds like what a rapist says because a rapist's whole thing is about bypassing consent... which yeah, is EXACTLY what statements like “It’s going to happen whether you like it or not” and "The inevitable march of progress" are doing.
If you are pushing past someone saying "No", it's worth taking a bit to look at why.
Make it practical for me to maintain my two bedroom apartment and food on the table without driving three (I have WFH two days) days a week and we can talk about me giving up the car.
In fairness - game companies can't lower the environmental impact, and AI companies aren't trying to. You can only change what you control, amd game companies can only control their use.
Our impression of tools is higly influenced by survivorship bias. Many dangerous tools have been banned, pulled from shelves by the manufacturer, restricted to experts, or even abandoned due to their nature. We just don't think about them due to their absence in our lives.
Seriously, Valve has been laying the ground work for this machine for YEARS, learning a lot along the way, but now they can't even budget properly because AI has allowed RAM producers to become a goddamn cartel!
If there is misdirected anger here, it's because there is a TON of anger that only an expert can properly work through it, and only through actual accountability we are not seeing could it remain directed where it belongs.
Honestly, I would like to challenge your premises here.
Any company that wants to stay arround has to charge at least the cost of doing buisness. These LLMs, on top of being heavily subsidized, have not had to take into account the cost of licencing. If we force big companies to pay licencing costs, they won't be able to continue selling LLMs.
Speaking of licencing costs: this only matters if copyright holders say "Yes". If they say "No", as they have a legal right to do so, then LLM companies start having an input problem.
Licencing costs also don't just matter during training. It's not at all unusual for licences to include royalties, meaning licence fees need to be paid out for the lifetime of the model, not just the training period.
SO lets say we started enforcing the law here:
Enfringing models would have to be retired, since many of the copyright holders would refuse to authorize use of their content.
LLMs get even more expensive, and companies do the math. They replace the LLMs with humans and things can chill some.
Ethical Developers no longer need to compete with GROK, since GROK's enfringement has caused it to be retired by legal means
The list of copywritten works allows work authors to skip proving their work was used in trainig, making their cases easier to build.
Then either the company should have the records to back the dates they chose to use, or someone failed to do their due dilligance before using the works for commercial purposes, and the company should be in trouble.
As my buisness does not depend on my lunch, this data was untracked. I could tell you what dependencies a piece of software had going back that far though.
Thing is, for something like a comment on the article, everything is "Reasonably Cited" - by which I mean if I quoted your quote, and someone had just the RSS feed entry of my quoting your quote, the author would be quickly discovered.
The law as described in the article is only asking companies to "identify and disclose copyrighted works", which could easily be done in the case of quoting a quote of a quote.
You would simply cite where you got it from and pass along relevant citations from where you got it from, referencing all relevant authors as impacted copyright holders. After all, the message containing the quote is a derivative work.
Sure, automate that fact-finding process for every single thing on the web. I’ll wait.
Well... it's probably doable anywhere that consented to be used for AI training. But you're right about it not being practical everywhere. Which IMO is an argument that said content should not be used in the first place. Our laws should not be written to protect "jerks [...] abusing long-standing social conventions in ways that are actively harmful" from the consequences of their failure to have common decency.
I don't see how this is a net negative?
Essentially California, a highly flammable state on regular water restrictions, would be loosing out on a high-heat, low-employment industry, and thus developers would be able to use the land for... anything else including space for wildlife?
The bill demands information that often does not exist, and cannot realistically be obtained.
Why can't it be obtained, Joe? WHY CAN'T IT BE OBTAINED?!
Look. I work with software. You give me any, ANY of our codebases and half an hour, and I will get you at least a shallow dependency list. Give me a day and I can script out some shit to process that into something actually usable.
You don't just shove things in your software and call it a day, developers HAVE A RESPONSIBILITY to manage the Other People's Work that is included in the final product.
And it's not just us either. I bet if I walked into a resturaunt and asked the owner "Who are your suppliers, you have one day to answer", they could get me a list of every supplier for a year or point me to the people who could. Hell, there are buisnesses where I could walk in with a product serial number and ask about the source of a specific washer used in the product, and they would be able to point me to the specific LOT that the washer came from.
So when I say I have ZERO sympathy for how hard it is to produce the list of worke that they used to train their products, I absolutely mean it.
So what if you don't know "Is this copywritten"?! If you didn't know then:
you should probably include it in your list
why the hell are you using it if you don't know its legal status
No seriously what the fuck were you thinking including a dependency you did not bother to understand the legal status of?
There is, what I will call, a minumum standard for doing buisness. If you don't meet that standard, you should not be doing buisness. Being able to answer "How do you source your inputs" is part of that minimum standard.
Glorbo?! After that whole Goncharov Incident?
The only way they can be silly is if they barely scratch the surface of what a technology nearly purpose built for scams can do...
If I remember correctly, the initial flagging of games came out of a legal concern. Players showed interest in knowing this information, so Valve put together a first pass on exposing it in the store. Which makes Valve even MORE in the right.
Let me make this clear as well. Consent, and violation of, is not a domain limited to sex crimes either. But we talk about it so little that we think sex crimes before we think about dinner tables or clothes.
I love how you ignored what I actually said to fight an argument I did not make ^.^
It's not their fault our society considers consent a sex thing. Stephen probably thinks it sounds like what a rapist says because a rapist's whole thing is about bypassing consent... which yeah, is EXACTLY what statements like “It’s going to happen whether you like it or not” and "The inevitable march of progress" are doing. If you are pushing past someone saying "No", it's worth taking a bit to look at why.
You give the industry where they bypass local regulations to use illegal gas generators too little credit for their efforts.
Make it practical for me to maintain my two bedroom apartment and food on the table without driving three (I have WFH two days) days a week and we can talk about me giving up the car.
In fairness - game companies can't lower the environmental impact, and AI companies aren't trying to. You can only change what you control, amd game companies can only control their use.
Our impression of tools is higly influenced by survivorship bias. Many dangerous tools have been banned, pulled from shelves by the manufacturer, restricted to experts, or even abandoned due to their nature. We just don't think about them due to their absence in our lives.
Seriously, Valve has been laying the ground work for this machine for YEARS, learning a lot along the way, but now they can't even budget properly because AI has allowed RAM producers to become a goddamn cartel! If there is misdirected anger here, it's because there is a TON of anger that only an expert can properly work through it, and only through actual accountability we are not seeing could it remain directed where it belongs.
Honestly, I would like to challenge your premises here. Any company that wants to stay arround has to charge at least the cost of doing buisness. These LLMs, on top of being heavily subsidized, have not had to take into account the cost of licencing. If we force big companies to pay licencing costs, they won't be able to continue selling LLMs. Speaking of licencing costs: this only matters if copyright holders say "Yes". If they say "No", as they have a legal right to do so, then LLM companies start having an input problem. Licencing costs also don't just matter during training. It's not at all unusual for licences to include royalties, meaning licence fees need to be paid out for the lifetime of the model, not just the training period. SO lets say we started enforcing the law here:
Then either the company should have the records to back the dates they chose to use, or someone failed to do their due dilligance before using the works for commercial purposes, and the company should be in trouble.
As my buisness does not depend on my lunch, this data was untracked. I could tell you what dependencies a piece of software had going back that far though.
Work Name: Copyright date-date. Public domain as of date.
Thing is, for something like a comment on the article, everything is "Reasonably Cited" - by which I mean if I quoted your quote, and someone had just the RSS feed entry of my quoting your quote, the author would be quickly discovered. The law as described in the article is only asking companies to "identify and disclose copyrighted works", which could easily be done in the case of quoting a quote of a quote. You would simply cite where you got it from and pass along relevant citations from where you got it from, referencing all relevant authors as impacted copyright holders. After all, the message containing the quote is a derivative work.
Well... it's probably doable anywhere that consented to be used for AI training. But you're right about it not being practical everywhere. Which IMO is an argument that said content should not be used in the first place. Our laws should not be written to protect "jerks [...] abusing long-standing social conventions in ways that are actively harmful" from the consequences of their failure to have common decency.I don't see how this is a net negative? Essentially California, a highly flammable state on regular water restrictions, would be loosing out on a high-heat, low-employment industry, and thus developers would be able to use the land for... anything else including space for wildlife?
- you should probably include it in your list
- why the hell are you using it if you don't know its legal status
- No seriously what the fuck were you thinking including a dependency you did not bother to understand the legal status of?
There is, what I will call, a minumum standard for doing buisness. If you don't meet that standard, you should not be doing buisness. Being able to answer "How do you source your inputs" is part of that minimum standard.