The rise of artificial intelligence (AI) has altered the technological landscape in ways that were unfathomable a few decades ago. Algorithms like GPT-3, created by OpenAI, can generate mind-boggling text that mimics human writing, fueling the progress of AI technologies.
Generative AI, the technology behind these algorithms, is the latest trend among big tech companies. Recently, premium tech companies, such as Microsoft, Google, Meta (formerly Facebook), and OpenAI, have been pioneering the use of this technology for their services.
Alongside all the excitement and utility, this advancement presents a puzzling question regarding copyright infringement. If AI, through machine learning, can generate such content, to whom does the copyright belong or does a copyright even apply?
AI algorithms learn from vast amounts of data dumped onto them. The software feeds the AI with large datasets, which it uses as a source of learning and eventually creating new content.
For instance, Google's software can analyze a series of images and generate pictures of cats that never existed, Meta can generate entire virtual worlds, and Microsoft has its AI writing an array of articles. These companies, along with others, generate new data using AI, sparking the debate on copyright infringement.
What makes this topic confusing is that AI algorithms do not simply reproduce the data fed into them. Rather, they analyze this data, digest it, learn from it, and spit out new content.
That changes the scene considerably because, essentially, the AI is generating new content from existing content, a process not too different from how a human would use their knowledge to write a report or draw a picture.
If an artist's painting is used as data, the machine can then produce new pictures by taking inspiration from the original. It gets complex because these iterations seem to be original works by the AI and not mere reproductions of the initial input.
This raises the question of originality and copyright. If these python codes or articles that the AIs produce are genuinely new works, then how can they be considered infringing?
This conundrum has legal researchers, industry lawyers, and copyright experts. It challenges the traditional copyright law in a way that wasn’t previously conceived.
Copyright laws were designed to protect the rights of the creators. But with AI becoming the creator, does the law fully apply, or does it need a revamp to accommodate the possibilities of AI?
At present, copyright law assumes that every act of creation is a result of human will and autonomy, overlooking the possibility of AI-generated content.
Opinions are divided on how to evolve these laws. Some argue that copyrights can’t apply to AI creations since these entities can't own rights. Others want the proprietors of the AI to own the copyright of the work produced.
The current state of law seems to lean more towards the latter opinion. In the case where AI is used to generate music or literature, the copyright seems to be assigned to the proprietor of the AI, not the machine.
However, this leaves the door open for unprecedented legal predicaments. If the AI generates something by drawing inspiration from a copyrighted work, does that constitute a breach?
Moreover, the issue isn’t just about ownership but also about liability. If an infringement does occur, is it the AI, the guardian of that AI, or the original data provider who is liable?
To compound the issue, these AI aren’t only using one data source. When thousands, if not millions, of data points are ingested, can we trace back the instances of infringement?
The road of copyright law through the labyrinth of AI seems to be more complex than initially expected. Time will only tell if the solution lies in redefining the laws altogether or in navigating through the existing ones.
Until then, the debate rages on, and we can only observe and contemplate the role of copyright law in the era of generative AI.