Technology

Don't threaten me with a good time.

If I do a book report based on a book that I picked up from the library, am I violating copyright? If I write a movie review for a newspaper that tells the plot of the film, am I violating copyright? Now, if the information that they have used is locked behind paywalls and obtained illegally, then sure, fire ze missiles, but if it is readily accessible and not being reprinted wholesale by the AI, then it doesn't seem that different from any of the other millions of ways we use data in every day life. Just because a machine learned it instead of a human, I don't believe that it makes it inherently wrong. I am very open to discussion on this, and if anyone has a counter-argument, I'd love to hear it, because this is a new field of technology that we should all talk about and learn to understand better.

Edit: I asked GPT-4 what it thought about this, and here is what it said:

As an AI developed by OpenAI, I don't access any copyrighted databases, subscription sites, confidential information, or proprietary databases for my learning or response generation. I was trained on a diverse range of internet text, which includes publicly available data in multiple languages and formats. The training also involves learning from large-scale datasets that contain parts of many kinds of texts (books, websites, scientific articles, etc.). However, I don't know specifics about which documents were in my training set or have access to any specific documents or sources.

Your question seems to pertain to how copyright law applies to AI. This is a complex and evolving area of law and policy, largely because AI systems like me do not "read" or "use" material in the same way a human would. A human might read a book and then consciously recall information from it when writing or speaking. In contrast, AI models do not have consciousness or awareness, they analyze statistical patterns in the data they were trained on and use these patterns to generate responses.

When a human uses copyrighted material to create something new, the law often considers this to be a derivative work. It is typically necessary to obtain permission from the copyright holder to create derivative works. However, AI doesn't use or recall copyrighted material directly; instead, it learns patterns from a huge amount of text data to predict or generate text based on a given prompt.

Dissemination of copyrighted material, such as through forum posts or Q&A platforms, without proper licensing or within the bounds of fair use, can be a violation of copyright law. AI doesn't 'disseminate' copyrighted material; it generates responses based on patterns it has learned.

That said, the legal interpretation of how copyright law applies to machine learning and AI is still evolving, and there can be a lot of complexities and nuances. Therefore, the information here should not be taken as legal advice.

As of my last update in September 2021, this was the case. However, I recommend checking the most recent legal standpoints and opinions, as this area is still evolving rapidly.

As an european, I'm ok with that. If they close, then I'll stick to using local LLMs then.

Copyright enforcement is for thee not for me.

Aw shucks.

You can read the actual proposal here - https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex%3A52021PC0206

The stuff in the article isn't a problem IMO, but the main issue is the huge amount of bureaucracy for smaller companies and initiatives.

Almost everything counts as "AI" :

(a) Machine learning approaches, including supervised, unsupervised and reinforcement learning, using a wide variety of methods including deep learning;
(b) Logic- and knowledge-based approaches, including knowledge representation, inductive (logic) programming, knowledge bases, inference and deductive engines, (symbolic) reasoning and expert systems;
(c) Statistical approaches, Bayesian estimation, search and optimization methods.

Read the whole thing. The reason OpenAI is opposing the law is not necessarily copyright infringement.

One provision in the current draft requires creators of foundation models to disclose details about their system’s design (including “computing power required, training time, and other relevant information related to the size and power of the model”)

This is the more likely problem.

Oh darn

The EU's stance is concerning. Their coming laws would benefit unlawful AI devs backed by dictatorships. (Edit: They'll do whatever they want to research and build more powerful AIs while devs in EU struggle due to heavy restrictions.) Currently, big techs are still learning how to build strong AIs, and giving dictatorships huge advantage like this is dangerous.