Platforms are selling your work to AI vendors with impunity. They need to stop | Hacker News Reader

Platforms are selling your work to AI vendors with impunity. They need to stop

werd.io

40 pointsbenwerd19 comments

I can't speak for anyone else, but I want to put my personal viewpoint forward.

If I answer something on a forum, or otherwise share my work online. Usually the social contract is: this is work someone did, in a context, with attribution. Attribution and link to the context is implied. If it's a toy example or teaches a pattern like how to use an interface it's usually implied that the example is too obvious to someone who is skilled in that art (using the interface) and is thus of insufficient novelty to qualify for copyright coverage. However the utility of that work as educational material, and promotional material for the creator, is worthy of the social contract of social credit / attribution.

After thinking it over, I'm leaning towards this stance. Does the entity understand the work enough to make an assessment of the following: 1) Are they proficient in the art (skill) at hand? 2) Do they rate the results as sufficiently unique to qualify for copyright rather than obvious to a skilled practicer in that field? 3) Are they an entity which is legally allowed to claim copyright? - If all three are yes, then MAYBE they can claim copyright for the results. As far as the forums and places with the examples; they should remain in the context they were created. Usually publicly accessible examples open to all without charge.

So any text/image which is publicly available, now can be used to generate a potentially very similar work (with AI tools which scraped it) without any attribution to its original author. Without any licensing limitations.

Is this how it works now?

Yup. As someone who put a lot of new information on the internet, I find it dispiriting. I'm happy that knowledge spreads further than expected, but I'm not happy that others are taking the reward while stripping me of even acknowledgement.

It feels like instead of fixing bicycles for my community, I'm sending them to a warehouse across the ocean for resale.

Even if you could be acknowledged it would be meaningless in the sea of a million attributions. It's really a problem that needs to be solved at the scraping step I feel. That the internet is "public" and thus scraping it is fine has always seemed a little bit too generous to me.

It's my work, on my server, I allow people who want to look at it do so, I never said OpenAI the company was free to connect to my sever and scrape it.

I agree with you. My biggest gripe with AI is that they just scrape and use my content without asking. As the license holder I should get a say in this, and the default answer is no.

But once again the tech industry shows a terrifying lack of understanding of consent. I wouldn't let these people anywhere near my drink.

It's not that they don't understand, it's that they don't give a shit.

Practically speaking, even if there could be some nuance to the law, that is what is actually happening. Even if the US passes laws, the rest of the world will crack on.

I don't post creative work publicly anymore, which means no one sees it. But barely anyone was going to anyway, and at least my work isn't getting exploited by corporations.

I am not against my work being part of a model that is for and used by society in principle. I am just against corporations taking from and exploiting the society that supports them.

I give my comments out to the world assuming a cc0 license.

I want AI vendors to use them, just like anyone.

I don’t like the idea of sites selling my comments to AI vendors. The recent Reddit deal where they are getting $60M for selling content freely given by users like me is more offensive than people scraping and doing stuff with it.

It’s one thing to scrape copyrighted info, but community generated content is supposed to be free.

Forgive me if this sounds callous, but how can users sign up for a platform, agreeing to a ToS that grants these platforms license to do whatever they want with your content, and then get angry when they do something with it?

It will be really interesting if AI starts stepping meaningfully into the music industry. We have really thoroughly explored copyright law for music in the courts. There are fewer gray areas for AI companies to hide in.

Many tech corporations appear to operate under the "its easier to ask forgiveness than to ask permission" mantra. So they simply don't care about copyright. They will happily break the law by downloading millions of copyrighted works, make billions from the results and then eventually pay back a few hundred million to those that it has wronged.

If Wordpress sites can be scraped directly, what additional benefit would there be to get the data through Automattic?

The benefit that you can’t sue openAI etc because you implicitly agreed to have your work scraped by them because you didn’t opt out in time.

I see a huge law suit steam rolling OpenAI's way. :)

> WordPress powers over 40% of the web, which, given these numbers, could lead to a significant payday for the company

What? 40% of the web is definitely not hosted by Automattic.

> But every public website can also be scraped directly by AI vendors

Exactly. So platforms are only facilitating the AI vendors' job by offering a bulk data transfer. There might be a indemnification clause somewhere which is probably covered by the platform's ToS extensively.

The issue is if AI vendors have a right to use public data to train their models the same way you train your brain by reading the content yourself. This will be settled in court soon enough, IMHO.

The article tries to cause outrage but lacks substance.

Right, not to mention that in many countries including EU countries, we already have laws that protect AI trainers. Google TDM exemptions. This isn't really an issue IMO; public data can already legally be used to train AI regardless of what platforms are doing with it.

I was skeptical of that 40% too so I went looking for the basis for it.

It's not totally unfounded: https://w3techs.com/technologies/overview/content_management estimates that WordPress is used by 63% of all websites that use content management systems, and they estimate 31% use no content management system at all, so it concludes that WordPress is used by 43% of all websites.

The thing is, though, this is not weighted by page views or time spent or any meaningful measure of usage. None of the top 20 websites use WordPress (https://www.statista.com/statistics/1201901/most-visited-web...) "WordPress powers over 40% of the web" misleads people into thinking it's 40% of everything people view, but I doubt that is true.

Lies, damned lies, and statistics!

What a perfect example. :-)

WordPress is an open source project that anyone can host.

That 40% number is probably true for WordPress installations.

Not all WordPress installations are hosted at Automattic (the company). Thus it's hard to understand how 40% of all websites using WP will give a huge payout to Automattic for sharing their customers' data (which is minor subset of all WP installations out there that make up that 40% number).

Reply on news.ycombinator.com