Patreon rejects "fair use" claims for AI training, calls for creator compensation

Alfonso Maruccia

Posts: 2,572   +956
Staff
LLM Paying Machine: Many AI startups and Big Tech players try to justify the theft of data and user-generated content used in LLM training as fair use. Patreon, a company designed to fairly compensate human creators, completely rejects the argument. AI corporations must pay up – something they are already doing for major journalism outlets – or face the consequences.

Jack Conte created Patreon to try and earn extra from his YouTube videos. The musician-turned-businessman is now managing a platform with 3 million monthly active users, and has plenty to say to big corporations operating chatbots and other AI platforms. First and foremost, these AI companies should stop crying foul and start paying content creators.

Conte talked about AI business ventures during a recent SXSW conference in Austin, Texas. He described LLMs as yet another transformative moment for computer technology, on par with major transitions such as going from downloading music on iTunes to streaming.

Change is not inherently bad, Conte said, adding that artists will survive the chatbot revolution and even thrive in the future. However, AI-focused corporations are doing something that the Patreon founder doesn't like at all: racking up massive troves of data to train or fine-tune their language models, without providing any compensation for the people who created it in the first place.

OpenAI and other companies, including tech giants such as Microsoft and Google, have traditionally tried to justify this content theft as fair use. Conte thinks the argument is "bogus" because corporations are signing multi-million dollar deals with major rights holders and publishers. Disney, Condé Nast, Vox, and Warner Music have already secured their "fair" compensation from OpenAI and other AI ventures.

"Why pay them and not creators – not the millions of illustrators and musicians and writers – whose work has been consumed by these models to build hundreds of billions of dollars of value for these companies?" Conte said.

The Patreon CEO is clearly trying to join the fray with his platform by allowing the community of writers, artists, and programmers to capture some of the AI-derived compensation. Conte confirms that he has not chosen an anti-AI stance because this would amount to being against technology or change.

Change happens either way, and Conte thinks LLMs and chatbots are here to stay. However, we should think about a future where artists can still be compensated while AI's text prediction machine ablates semantic nuance. In the end, the Patreon founder thinks that humans will keep enjoying other humans' work and artistic expression for the foreseeable future – no matter how complex and convincing AI lies become.

Permalink to story:

 
It's already too late for it. Ai is in such a paste it will revolutionary change everything. Likely even bigger then the whole industrial time back in the 1900's. You wait.
 
Current copyright law appears incompatible with the capabilities of artificial intelligence (AI), forcing lawmakers into a difficult choice about its future. A public opinion poll could provide valuable insight into this issue.

Essentially, there seem to be two primary options. The first is to abolish copyright entirely and put all materials and work in public domain and introduce tax incentives as compensation, such as income tax and VAT exemptions, for trading materials in the public domain (e.g., allowing anyone to reproduce and sell a Harry Potter book without these taxes once it's public domain). The second option is to prohibit the development and use of AI.

Lawmakers must choose between these two paths; this fundamental conflict cannot be avoided. Currently, individuals can generate content using AI and present it as their own original work. AI systems produce output at a rate orders of magnitude greater than humans, and with higher quality (that’s why they choose to use it).

The transformer architecture, a common design in modern AI, employs a highly interconnected 'all-to-all' structure. This differs significantly from the human brain's more localized processing and attention mechanisms. In essence, AI can connect vast amounts of information in ways that surpass human cognitive capacity, making its capabilities effectively superhuman.

Therefore, the challenge is to decide which system to prioritize: the 'superhuman' capabilities enabled by AI, or the limitations and established norms of human creativity and ownership as currently understood.
 
Current copyright law appears incompatible with the capabilities of artificial intelligence (AI), forcing lawmakers into a difficult choice about its future.
Current copyright law still works here. Artists, painters, songwriters, novelists -- all learn their skills by studying the works of past masters and currently popular works. If they duplicate a copyrighted work, they've infringed. If they merely use what they've learned to produce a competing work, they haven't. AI is doing the same ... just faster.

Forgetting AI entirely, the only reason sites like TS exist (along with 99% of the Internet's current content) is because humans read some already-published copyrighted work, paraphrase it into their own words, and repost. I don't see that AI doing the paraphrasing changes the legalities whatsoever.
 
Back