Scientists Debate the Ethics of AI and Copyright Infringement
Artificial intelligence (AI) has captivated human interest and fear for many years. While people may worry that AI could eventually outsmart humanity and threaten our existence, current AI iterations lack self-awareness and can only predict text based on the data they’ve studied. However, the realism of AI-generated text can be quite convincing.
To understand how AI works, it’s more efficient to analyze its output data using probabilities rather than trying to penetrate its psychology. In fact, this approach has already yielded results. A team of scientists from the University of California, Berkeley, conducted research on how AI decides what to display to users based on the data it receives at entry. By analyzing how AI responds to user requests, the scientists were able to draw conclusions about the data used for training. They discovered that modern generative models were trained on copyrighted works, particularly the ChatGPT model. This evidence can be particularly concerning for scientists, who are increasingly pushing for more open data sets to train neural networks to create less biased final models.
However, a ban on the use of copyrighted works in chat bots isn’t necessarily a solution either. Although chat bots don’t necessarily copy an entire original work, they do create a mathematical set of connections that mix the work with other data. Even if the final result doesn’t appear to replicate the copyrighted work, the copyright owner may still argue that the work is a derivative of their intellectual property, which was created without their permission.
Moreover, modern language learning models (LLMs) are highly creative, able to learn and improve quickly. Legal barriers could hinder their developments in the future, which might mean we have to decide between legal constraints on their development or a quicker path that allows them to evolve without such restrictions. Perhaps the public will choose to embrace neural networks despite the issues of copyright infringement, especially if they continue to make daily tasks more efficient.
Despite our worries, AI still has a long way to go before it can challenge humanity’s existence. However, as technology advances and AI becomes more sophisticated, it’s important to thoroughly consider what ethical and legal implications follow their integration into our daily lives.