New certification for ethics in artificial intelligence
Due to the need to regulate the content generated by artificial intelligence systems, today a proposal called Fairly Trained arises, an initiative led by Ed Newton-Rex, former executive of Stability AI, which seeks to certify GenAI companies that use licensed training data, ensuring the consent of the creators of such content.
One point to highlight because Fairly Trained is emerging is the absence of certifications for text generation models. According to Newton-Rex, these (text generation) models do not meet the criteria because they rely on a large volume of copyrighted content.
On the other hand, OpenAI, the company behind ChatGPT, has commented that it is virtually impossible to develop such models without using this type of content, underscoring a central dilemma in the GenAI field. Fairly Trained's initiative is a first step towards striking a balance between technological innovation and respect for intellectual property.
The road to fully copyright-respecting AI models is fraught with obstacles, this certification poses a viable alternative and one that aligns with Newton-Rex's vision of less data-trained language models under appropriate licenses, which would open a window of hope towards a more ethical future within AI.
The Fairly Trained initiative signals a paradigm shift in how industry and consumers perceive and use artificial intelligence and could drive the development of GenAI with ethical practices, as well as the potential to give pre-created content the importance it is due.
The project is based on several points such as Corporate Awareness and Responsibility: which could differentiate companies from those that are not certified, Consumer Preferences: certification can be a deciding factor in purchasing, Innovation in AI Model Training: research and development of new AI training methods that do not rely on large amounts of copyrighted data could be encouraged, which means innovation in the field, and finally, Awareness and Education: the opportunity to educate the public about copyright and ethical use of data in AI.
It is expected that this project will have continuity, even when faced with factors within the industry such as adoption and implementation by the creative companies, what cannot be denied, is that it has the potential to positively influence the ethical and responsible development of generative artificial intelligence.
All the details at https://www.fairlytrained.org/
