The Challenges of Defining Human Content in AI
Amidst the rapid advancements in AI technology, there’s growing concern about how to treat human “content.” Societies must closely examine this issue by first defining what constitutes human content and broadening this category of information. Human content encompasses creative works such as songs, poems, and visual art, as well as professional intellectual property, including information about an individual’s job routines, strategies, and expertise. Additionally, it includes personal attributes like likeness, characteristics, voice, face, and body – essentially, what makes a person unique.
All of this personal data deserves protection. The European GDPR compliance cases serve as a warning sign for the complexities ahead. The issues surrounding intellectual property and fair use in AI extend far beyond current legal frameworks. Leaders are discussing three critical aspects to determine appropriate AI regulation.
Influence and Ownership
One of the biggest questions is determining when an AI system crosses the line from general influence to content theft. AI companies might argue that their systems gather information piecemeal from various sources, potentially siphoning off underlying intellectual property. Lawsuits from news organizations like the NYT against AI model companies illustrate this concern.
A conversation between Chris Anderson and Sam Altman of OpenAI highlights this issue. Altman discussed how AI can take direct influence or coalesce from broader training sets, making it difficult to discern the difference. Anderson questioned whether this constitutes IP theft, sparking a discussion about attributing sources and dividing compensation.
The consensus is that it’s challenging to determine when the system crosses the line. Syed Balkhi notes that AI training aims to teach pattern recognition and generate outputs mimicking human creativity, raising critical legal questions about copyright infringement versus fair use. AI advocates argue that training data is used transformatively, while critics see it as exploitation of copyrighted material without creators’ consent.
Consent for AI Use
Consent is another key issue. Demonstrations of AI applications often involve using a host’s data without explicit consent, prompting concerns about how to enforce consent. This is a question that requires careful consideration.
AI Evaluation and Promotion
The discussion between Altman and Anderson also touched on AI’s potential to review and promote creative work. In the movie “Her,” an AI evaluates someone’s work and influences decisions about sharing it with the world. This concept is both intriguing and counterintuitive, as AI acts as an agent for our collective consciousness, using internet information to evaluate and recommend content.
Artificial General Intelligence (AGI)
AGI systems will be able to mine professional data to replicate roles or perform tasks on our behalf. This raises concerns about job displacement and the need for a new regulatory framework. Two themes emerge: AGI can ‘do my job’ or ‘do stuff for me,’ from replicating professional roles to performing undesirable tasks.
As we navigate these complexities, it’s essential to continue thinking about the implications of AI and AGI on our society, including the need for clear regulations and protections for human content.
