Posted by geox 17 hours ago
The licensing standard they're talking about will achieve nothing.
Anti-bot companies selling scraping protections will run out of runway: there's a limited set of signals, and none of them are robust. As the signals get used, they're also getting burned. And it's politically impossible to expand the web platform to have robust counter-abuse capabilities.
Putting the content behind a login wall can work for large sites, but not small ones.
The free-for-all will not end until adversarial scraping becomes illegal.
Well, with one alternative: Edge.
Syndication is the answer. Small artists are on Spotify, small video makers are on YouTube.
See the problem?
user queries "static" training data in LLM; LLM guesses something, then searches internet in real-time for data to support the guesses. This would be classified as "browsing" rather than trawling.
(the searched data then get added back into the corpus, thus sadly sidestepping all the anti-AI trawling mechanisms)
Kind of like the way a normal user would.
The problem is, as others have already mentioned, how would the LLMs know what is a good answer versus a bad, when a "normal" user also has this issue?
I think a big difference is that there’s no micro transactions or compulsory licensing for content, so it always feels patently unfair to buy a subscription to read one article.
You can see artifacts when their servers are at queue load and you see the URLs, a few resources have the JWT with the account details in the URL. IIRC the clearname of the account in the token is Masha Rabinovich, with an email account masha@dns.li, an identity that has cropped up in various investigations [1][2].
[1] https://gyrovague.com/2023/08/05/archive-today-on-the-trail-...
[2] https://webapps.stackexchange.com/questions/145817/who-owns-...
Those things were afterthoughts because for the most part the experimental methods sucked compared to the real thing. If we were in mid 2016 and your LSTM was barely stringing together coherent sentences, it was a curiosity but not a serious competitor to StackOverflow.
I say this not because I don’t think law/ethics are important in the abstract, but because they only became relevant after significant technological improvement.
While this likely has no legal weight (except for EU TDM for commercial use, where the law does take into account opt-outs), they are betting on using services like CloudFlare and Fastly to enforce this.
[1] https://www.investors.com/research/the-new-america/reddit-st...
I wouldn't be quite so sure about that. The AI industry has entirely relied on 'move fast and break things' and 'old fart judges who don't understand the tech' as their legal strategy.
The idea that AI training is fair use isn't so obvious, and quite frankly is entirely ridiculous in a world where AI companies pay for the data. If it's not fair use to take reddit's data, it's not fair use to take mine either.
On a technological level the difference to prior ML is straightforward: A classical classifier system is simply incapable of emitting any copyrighted work it was trained on. The very architecture of the system guarantees it to produce new information derived from the training data rather than the training data itself.
LLMs and similar generative AI do not have that safeguard. To be practically useful they have to be capable of emitting facts from training data, but have no architectural mechanism to separate facts from expressions. For them to be capable of emitting facts they must also be capable of emitting expressions, and thus, copyright violation.
Add in how GenAI tends to directly compete with the market of the works used as training data in ways that prior "fair use" systems did not and things become sketchy quickly.
Every major AI company knows this, as they have rushed to implement copyright filtering systems once people started pointing out instances of copyrighted expressions being reproduced by AI systems. (There are technical reasons why this isn't a very good solution to curtail copyright infringement by AI)
Observe how all the major copyright victories amount to judges dismissing cases on grounds of "Well you don't have an example specific to your work" rather than addressing whether such uses are acceptable as a collective whole.
Sure they do. Every time a bot searches, reads your site and formulates an answer it does not replicate your expression. First of all, it compares across 20.. 100 sources. Second, it only reports what is related to the user query. And third - it uses its own expression. It's more like asking a friend who read those articles and getting an answer.
LLMs ability to separate facts from expression is quite well developed, maybe their strongest skill. They can translate, paraphrase, summarize, or reword forever.
> Every time a bot searches
We are talking about LLMs by themselves, not larger systems using them.
> LLMs ability to separate facts from expression is quite well developed
It is not. Whether you ask an LLM for an excerpt of the bible, or an excerpt of The Lord of the Rings, the LLM does not distinguish. It has no concept of what is, and what is not, under copyright.
> Observe how all the major copyright victories amount to judges dismissing cases on grounds of "Well you don't have an example specific to your work" rather than addressing whether such uses are acceptable as a collective whole.
Well, all a judge can/should do is to apply current law to the case before them. In the case of generative AI then it seems that it's mostly going to be copyright and "right of publicity" (reproducing someone else's likeness/voice) that apply.
Copyright infringment is all about having published something based on someone else's work - AFAIK it doesn't have anything to say about someone/something having the potential to infringe (e.g. training an AI) if they haven't actually done it. It has to be about the generated artifact.
Of course copyright law wasn't designed with generative AI in mind, and maybe now that it is here we need new laws to protect creative content. For example, should OpenAI be able to copy Studio Ghibli's "trademark" style without requiring permission?
This is true, and I do not mean to suggest it is bad. But rather, that it leaves uncertainty. These cases can all be struck down without reducing the possibility that if one does stick, the entire industry is at stake.
> Copyright infringment is all about having published something based on someone else's work - AFAIK it doesn't have anything to say about someone/something having the potential to infringe (e.g. training an AI) if they haven't actually done it. It has to be about the generated artifact.
A notable problem here is that AI models are not "standalone products" but tools provided as a service. This complicates the situation.
Take Disney/Universal's case against Midjourney, which is both about the models but also the provision of services.
Even if only the latter gets deemed illegal, that's ruinous for the big AI companies. What good is OpenAI if they can't provide ChatGPT? Who would license a LLM if the act of using it creates constant legal risks?
If this intended to refer to Judge Alsup, it is extremely wrong.
A “classical” classifier can regurgitate its training data as well. It’s just that Reddit never seemed to care about people training e.g. sentiment classifiers on their data before.
In fact a “decoder” is simply autoregressive token classification.
The right thing would be for the end users to receive the compensation Reddit is getting from AI companies.
Is there even one example of a “tech mega corp” that has grown to control more than 1/5 of its market without this circling back to hurt people in some way? A single example?