One of the biggest issues with content creation has always been the issue of originality which legally revolves around intellectual property rights. The forms of contents we refer to in this discourse are audio contents, visual contents and audio-visual contents. The significance of the issues raised here cannot be overemphasized because content is king not only to marketers but for professionals. Originality here refers to the uniqueness of contents; produced without being an imitation of another person’s work. It is important to point here that, producing original contents is good and establishes authenticity and uniqueness. However, originality of contents is arguably not easily achievable and it is sometimes very difficult to determine if a content is original or not. Reason being that the way around content creation has drastically been revolutionized by the developments of technology; particularly the feats recorded with AI in this digital age.
Let us imagine a room filled with an endless supply of building blocks. Each block represents a piece of content, and technology is the tool that allows us to create new and unique structures with these blocks. In the past, creating content was like having a limited supply of building blocks, making it difficult to construct anything truly original. However, with technology, the number of building blocks available has increased, allowing us to build new and innovative structures with ease.
Just as a room filled with building blocks can lead to an abundance of creativity and new ideas, the increase in technology has opened up a world of possibilities for content creation. With tools like video editing software, graphic design tools, and AI-powered content creation, we can now create what people may argue to be original and engaging content on a massive scale. However, just as a room filled with building blocks can become clustered and overwhelming; the sheer volume of content being produced can sometimes make it difficult to identify unique and original.
Generative AI refers to AI systems that are capable of generating content, such as text, images, audio or video; making up the three forms of contents highlighted in the earliest part of this discourse. Generative AI systems work from learning data from large amounts of existing data, to using that knowledge to generate new content. For example, a generative AI system that has been trained on a large dataset of text might be able to generate new articles, or even entire books.
There are a variety of different approaches to generative AI, including deep learning algorithms like Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), as well as more traditional rule-based systems. Generative AI has the potential to revolutionize the way we create and consume content. For example, generative AI could be used to automate the creation of content and free up time and resources for other human tasks.
In all these, we should not discount the potential misuse of generative AI for creating contents as well as the legal and ethical concerns around its use. There is the possibility that the information given out by an AI may not be correct and credible, reason being that the quality of content generated by AI is only as good as the input data of the AI system. Let us avert our mind to the age-long computer science and information technology GIGO acronym which stands for “Garbage In, Garbage Out”. The implication of the GIGO issue with generative AI is that if the input data in which the AI system is trained is of poor quality or contains errors, the output produced by the system will not be immune to the error and is going to be of poor quality too.
For instance, a good use and review of one of the most trending innovations of 2023 which is ChatGPT clearly underlines how so much can be done with AI for content creation while also highlighting the GIGO issue with AI. ChatGPT is a language model that uses a type of artificial neural network called a transformer to generate human-like text prompts. When a user inputs a prompt or question, ChatGPT analyzes the input and generates a response based on the patterns it has learned from the text data it was trained on. This makes ChatGPT useful for a wide range of applications, such as language translation, chatbots, and content creation. ChatGPT does not generate original contents but its response can be used for content creation.
In our interview with ChatGPT, one of our correspondent literally accused ChatGPT of copying data from the internet. ChatGPT was overwhelmed in shock and quickly asserted that it does not directly copy data from the internet. According to the AI system, it is trained on large corpus of text data, which includes text from variety of sources, such as websites, books and other documents available on the internet. However, it is important to point out here that there is the possibility that most of the input data with which ChatGPT was and is being trained on are mostly internet materials.
In this digital age, the internet can be seen as a global basket of data; it is a vast network of interconnected computers and servers that host an enormous amount of data in various forms, such as text, images, videos and audios. This data is constantly being generated, shared, and accessed by billions of users around the world. Significantly, the internet has become the primary source of information and communication for people around the globe, providing access to a wealth of knowledge, ideas, and perspectives. It has transformed the way we access and share information, breaking down geographical barriers and making it possible to connect with people and organizations from anywhere in the world.
There is no unified control over the internet, though there are some organizations and bodies like the Internet Engineering Task Force (IETF) and the Internet Corporation for Assigned Names (ICANN) that plays a role in regulating and governing it. Since the internet is a global basket, everyone can put data in it; except for persons who are restricted from accessing it. As a result of the open access of the internet, we can find good and credible, and also bad and misleading data littered everywhere. So when ChatGPT says it has been trained with data from the internet, it implies that if its creators have fed it with some erroneous data, the output is going to be erroneous too.
Moving on, it is important to point out that there are some AI generative systems that generate their output data without directly replicating or giving out already existing data from the internet; at least to the lay man’s knowledge of the users. At this point, we are going to be making reference to AI generative tools for images. Presently, we have several AI generative tools that are capable of generating realistic images of people who do not exist in real life and original artworks. To mention just a few, there is ThisPersonDoesNotExist which uses Generative Adversarial Network (GAN) to generate images of people who do not actually exist. GPT-3 which is currently the most advanced version of ChatGPT is also capable of generating realistic images of people, animals, and objects.
For this article, we also interviewed Ejiroghene Malcolm Obiuwevbi, a young full stack developer in Nigeria, who has also built an AI image generator. According to Ejiroghene, his image generator is an application that creates images on the fly based on set of parameters; like the texts a user inputs describing the image he needs to be generated and the desired size of the image. On whether or not, the image generator just replicates or gives its output data from the internet, he stated that the images are created in realtime by the AI and they are copied from anywhere; “what the AI does is read the description of the image given by the user and generates the image with 99.9% accuracy”.
We had a test run on the image generator and to the best of our knowledge and usage; the AI application actually works to generate original and “tear rubber” images. One thing that noticeably stands out is the fact the AI image generator does not generate the same image more than once, even if it is the same description parameter. Ejiroghene made us to understand that his drive behind building the image generator was his personal experiment on OpenAI’s DALL.E algorithm which can create realistic images and arts from a description in natural language. According to him, he found OpenAI’s project very fascinating and he decided to do a personal project after seeing a tutorial on it. For his futuristic projections, he believes so much will be done with AI generative tools and for his image generator, he believes and it can be used to generate gravatars for game characters amongst other things.
With regards to the intellectual property rights issues with AI generated contents, it is important to point out here that the ownership rights of contents created by AI varies depends on the system. For an AI system like ChatGPT, the fact that a book or an original material written by a human author is used in training the AI, ownership of such data would still be with the original author. In fact, it is very necessary that the developers of AI systems like ChatGPT integrate features that would allow the referencing of materials with which the responses given are based. When AI generated systems produce data as responses to the users, it is very imperative that the users exercise care in using those data so as not to infringe on intellectual property rights.
Currently, it is very difficult to lay proprietary claims on AI generated contents because the data that have been used in training the AI systems are so much that only the developers of the systems may easily know the sources from where they collated data. It is important to point out here that even ChatGPT exempts itself from liability for any intellectual property rights infringement that may arise as a result of the responses it gives. In fact, ChatGPT believes that once a user inputs any copyrighted material into it and the model generates a response that includes the copyrighted material, the user, and not ChatGPT, would be held responsible for any infringement.
No amount of exemption clauses can comfortably settle the intellectual property rights issues that are going to arise as a result of the usage and ownership of data or contents generated with AI. Unfortunately, when AI is built and used to generate contents and information, except they are so trained to understand and comply with applicable laws and regulations governing intellectual property rights, there is the likelihood that we may witness a lot of court cases and actions on this matter. Right now, many people are so intrigued and excited about the potentials of generative AI. However, when the excitements are down, people may start to notice the several replications in contents generated with AI as a results of the increased usage, others may also start to lay claims to some of the contents generated. This is an inevitable projection because human beings are gradually becoming so dependent on AI powered systems for their work.
A step to solving some of the intellectual property rights issues with generative AI is direct referencing. The creators and promoters of generative AI should work towards training the AI to include possible references to contents from where the responses and output data are given. This will definitely not make the AI systems less useful but assist users to give credits or even seek permissions; when necessary, before using the output data.
Importantly, the issue of over-dependency on AI systems by human being is arguably not a very bad thing, because AI literally makes work easy for us. However, users of AI should also understand that they also have a responsibility when using generative AI for content creation. Even when the output data looks original, users should learn to scrutinize and make modifications, rather than just copying and pasting the output data. AI systems are created to help and make our brain work easy and efficient as human beings and not to replace our brain. Presently, we have people creating audio and visual contents with AI, and even writing as much as a full book with AI. While we congratulate them on whatever magic they can create with AI and applaud them for the adoption, we must advise that all users exercise care when using these tools.
To conclude this discourse, it is important that we mention that while it is important to consider the ethical and intellectual property rights issue with the use of generative AI, we must understand that the world is a big place and human beings are naturally creative. There is the possibility that there may be coincidence. ¯In this regards, the creator may not even be aware that what he thought was an original and unique idea has already been similarly created by someone else. This is a very complex intellectual property rights issue and can be very difficult to solve. In all of these, there is so much that is yet to be done and it may take a while for us to have settled legal principles and frameworks regulating the innovations men are creating. There are going to be so much of legal issues for lawyers and the courts in different legal systems of the world to determine.
