Summary of TED Talks Daily Episode: "How AI Models Steal Creative Work — and What to Do About It" by Ed Newton-Rex

Release Date: March 14, 2025

In this compelling episode of TED Talks Daily, host Elise Hu introduces Ed Newton-Rex, a renowned generative AI expert, who delves into the ethical and legal dilemmas surrounding the use of creative works in training artificial intelligence models. Newton-Rex articulates the conflict between AI development and the rights of creators, advocating for a more equitable and sustainable relationship between the two domains.

The Core Issue: Unlicensed Use of Creative Works

Newton-Rex begins by highlighting the unethical practices prevalent in the AI industry, where companies utilize vast amounts of creative content without obtaining proper licenses or permissions. He identifies the three critical resources AI companies require to build their models: people (engineers), compute (GPUs), and data. While substantial investments are made in the first two, the third resource—training data—is often sourced for free, primarily through unauthorized scraping of copyrighted material.

“AI companies spend vast sums on the first two [people and compute]. Sometimes a million dollars per engineer and up to a billion dollars per model. But they expect to take the third resource, training data, for free.”
(02:37)

Widespread Use of Copyrighted Material

Newton-Rex presents evidence of the widespread use of copyrighted works in AI training. Referencing a study by the Mozilla Foundation, he notes that 64% of 47 large language models released between 2019 and 2023 were trained using datasets like Common Crawl, which include numerous copyrighted materials. An additional 21% of these models lacked transparency, making it difficult to ascertain their data sources.

“Training on copyrighted work without a license is rife... training on copyrighted work without a license has rapidly become standard across much of the generative AI industry.”
(05:15)

Negative Impacts on Creators

The unauthorized use of creative works has tangible negative consequences for creators. Newton-Rex explains that generative AI inherently competes with its training data, leading to reduced income and opportunities for artists. He shares poignant examples, such as Kelly McKernan from Nashville, whose income dropped by 33% almost overnight after an AI image model trained on her works gained popularity.

“Kelly's income fell by 33% almost overnight. Illustrators around the world report similar stories being outcompeted by AI models they have reason to believe were trained on their work.”
(10:50)

Furthermore, platforms like Upwork have observed an 8% decline in demand for freelance writing tasks, increasing to 18% for lower-value tasks, attributable to the rise of generative AI.

Legal Ambiguities and Fair Use Debate

A significant portion of the discussion focuses on the legal gray area surrounding AI training on copyrighted material. While AI companies often cite the Fair Use exception, creators vehemently disagree, arguing that mass exploitation of their work goes beyond the intended scope of this legal provision.

“Creators and rights holders strongly disagree, saying there's no way this narrow exception can be used to legitimize the mass exploitation of creative work to create automated competitors to that work.”
(08:22)

With approximately 30 ongoing lawsuits challenging AI companies, the legal landscape remains uncertain, leaving creators in a precarious position.

Proposed Solution: Licensing Training Data

Newton-Rex advocates for a licensing framework as a viable solution to balance the needs of AI developers and creators. He emphasizes that licensing is a standard practice in other industries and can be effectively applied to AI training data. Despite AI companies' claims about the impracticality of licensing due to the sheer volume of data, Newton-Rex counters that numerous negotiations and models, such as revenue sharing, have already proven successful.

“There are multiple options available to you if you want to build your model without infringing copyright... multiple companies are doing it already.”
(13:45)

He cites the example of Stability AI, which released an AI music model trained exclusively on licensed music, demonstrating the feasibility of ethical AI training practices.

Public and Creator Support for Licensing

Public opinion strongly supports the notion that AI companies should seek permission and compensate creators for using their works. Newton-Rex references a poll by the AI Policy Institute, revealing that 60% of people oppose AI training on publicly available data without consent, and 74% believe AI companies should compensate data providers.

“74% said yes [AI companies should compensate data providers] and only 9% said no.”
(15:30)

Additionally, a burgeoning movement among creators is taking shape. Newton-Rex mentions the launch of a statement on AI training, an open letter signed by over 11,000 creators, including Nobel laureates and Academy Award winners, condemning the unlicensed use of their works.

“These artists, these creators, view the unlicensed training on their work by generative AI models as totally unjust and potentially catastrophic to their professions.”
(16:45)

A Path Forward: Symbiosis Between AI and Creativity

Concluding his talk, Newton-Rex envisions a future where generative AI and human creativity coexist harmoniously. He stresses that mutual respect and fair compensation are crucial for this symbiosis to materialize. By licensing training data, AI companies can continue innovation without undermining the livelihoods of creators or disrupting the creative ecosystem.

“A future in which generative AI and human creativity can coexist not just peacefully, but symbiotically... it's not too late to change course.”
(17:20)

Newton-Rex calls for collective action, urging AI companies to follow the lead of pioneers like Stability AI, encouraging employees to advocate for fair practices, and inspiring consumers to question the origins of AI models they utilize.

Conclusion

Ed Newton-Rex’s insightful analysis underscores the urgent need for ethical standards in the development of generative AI. By highlighting the detrimental effects of unlicensed use of creative works and proposing actionable solutions through licensing, Newton-Rex provides a roadmap toward a more equitable and sustainable integration of AI into the creative industries. His call to action emphasizes that fostering a respectful and mutually beneficial relationship between AI developers and creators is not only possible but essential for the future of both fields.

Notable Quotes:

• “AI companies spend vast sums on the first two [people and compute]. Sometimes a million dollars per engineer and up to a billion dollars per model. But they expect to take the third resource, training data, for free.”
  (02:37)

• “Training on copyrighted work without a license is rife... training on copyrighted work without a license has rapidly become standard across much of the generative AI industry.”
  (05:15)

• “Kelly's income fell by 33% almost overnight. Illustrators around the world report similar stories being outcompeted by AI models they have reason to believe were trained on their work.”
  (10:50)

• “Creators and rights holders strongly disagree, saying there's no way this narrow exception can be used to legitimize the mass exploitation of creative work to create automated competitors to that work.”
  (08:22)

• “There are multiple options available to you if you want to build your model without infringing copyright... multiple companies are doing it already.”
  (13:45)

• “A future in which generative AI and human creativity can coexist not just peacefully, but symbiotically... it's not too late to change course.”
  (17:20)

This episode serves as a crucial conversation starter for anyone interested in the intersection of artificial intelligence and creative industries, urging stakeholders to consider the ethical implications and work collaboratively towards a fair and innovative future.