- Modern AI technology resembles the mythical birth of Athena, but it relies on existing content for creativity.
- Some AI companies justify using content without compensating creators, sparking ethical debates.
- Fairly Trained, led by Ed Newton-Rex, certifies companies adhering to ethical AI training practices.
- Nine companies, including Bria, have received certification for their ethical AI models.
- Bria asserts its image-making models are trained exclusively on licensed content from reputable sources.
- Fairly Trained seeks to promote responsible AI innovation while ensuring societal benefit.
Main AI News:
In the realm of modern technology, harnessing the capabilities of powerful text- or image-generating artificial intelligence can evoke imagery akin to witnessing the mythological birth of Athena. As if she were emerging fully formed and armored from the forehead of Zeus himself, these AI systems manifest lucid paragraphs or realistic images with lightning speed, captivating the imagination of those who behold them.
However, a parallel narrative to the Greek myth presents Zeus not as a creator, but as a regurgitator. In this version, he consumes his pregnant wife, Metis, who carries Athena and crafts her armor, only for Athena to spring forth from Zeus’s mind after Metis gives birth. Similarly, generative AI systems rely on pre-existing content for their creative output. They break down human-made content into digestible bits before weaving them together to generate new material. For instance, OpenAI’s GPT-3.5, which powers the popular ChatGPT, was trained on a vast corpus of around 300 billion words scraped from sources like Wikipedia.
While some AI companies justify training models on existing content without compensating human creators, others vehemently oppose this stance. Stability AI, the creator of the image generator Stable Diffusion, argues that AI development constitutes a transformative and socially beneficial use of protected content under fair use principles. This viewpoint has been met with contention, as exemplified by the New York Times’ lawsuit against Microsoft and OpenAI for allegedly using the newspaper’s stories without permission to develop chatbots.
The ethical considerations surrounding AI development have prompted individuals like computer scientist Ed Newton-Rex to take action. Newton-Rex departed from Stability AI to establish Fairly Trained, a nonprofit organization that certifies companies adhering to ethical training practices for their AI models. Fairly Trained aims to distinguish between companies that ethically source content for training AI models and those that do not, providing clarity in an increasingly complex landscape.
To date, Fairly Trained has awarded annual certification to models from nine companies, including Bria, an Israeli AI firm backed by $24 million in Series A funding. Bria asserts that its image-making models are exclusively trained on licensed pictures from reputable sources such as Getty Images. Newton-Rex elaborates on the criteria used by Fairly Trained to certify companies like Bria, underscoring the organization’s commitment to fostering ethical AI development practices.
Despite the challenges and controversies surrounding AI development, Newton-Rex remains optimistic about the future of AI. Through initiatives like Fairly Trained, he seeks to promote responsible innovation while ensuring that AI technology benefits society as a whole.
Conclusion:
The emergence of Fairly Trained and its certification process signifies a growing emphasis on ethical considerations in AI development. Companies like Bria, backed by substantial funding, stand to gain market trust and reputation by aligning with ethical training practices. However, those neglecting ethical standards risk facing legal and reputational challenges, potentially hindering their market viability in the long run. As consumers increasingly prioritize ethical considerations, adherence to fair training practices is poised to become a critical differentiator in the competitive AI market.
