Legal issues involved around Large Language Model

In recent years, the world has seen an incredible rise in the development of Artificial Intelligence (AI) technologies. From voice assistants to self-driving cars, AI has become an essential part of our lives. One of the most prominent AI applications to emerge in recent years is the Large Language Model. From natural language processing (NLP) to machine translation, Large Language Models have a wide range of applications and have been adopted by many industry leaders. In this post, we will take a closer look at Large Language Models and delve into some of the legal implications that arise from their use.

How does the Large Language Model (LLM) work?

Large language models are computer models that use a large amount of text data to learn the relationship between words, phrases, and sentences in a given language. The models are trained by being exposed to large amounts of text data, usually from a variety of sources, such as books, newspapers, magazines, and online articles. The models are then able to generate accurate predictions of how words and phrases should be used in a given context. By having access to such a large amount of data, the models can understand the nuances of the language and its usage in different contexts. This can lead to more accurate predictions and better natural language understanding.

Who will be held responsible for such copyright infringement?

The issue of whether a model or the organization that created it can be liable for copyright infringement arises when models generate text that is similar or identical to existing copyrighted works. It is debatable whether it is possible to hold a model responsible in such a situation. Furthermore, it is uncertain if the generated text can be considered a derivative work and if the original copyright holder(s) would have any rights over it.

Can Artificial Intelligence (AI) be harmful to people?

Artificial Intelligence (AI) has caused harm to people in a number of ways. One of the most prominent examples of AI causing harm to people is when AI technology was used to automate decision-making processes. This has been seen in various contexts from financial lending to hiring decisions.

For example, in 2017, Amazon’s AI hiring system was found to be biased against female candidates. The algorithm was designed to sort through resumes and hire candidates with the highest scores, however, it was found to favour male candidates. The algorithm was quickly scrapped and Amazon had to start the hiring process again.

Another example is in financial lending decisions. AI algorithms have been used to automate decisions about who should be granted a loan. However, these algorithms have been found to be biased and discriminate against certain groups of people. For example, an AI loan decision system was found to be more likely to reject loan applications from African American people compared to white people.

Another example of where AI caused harm to people was when an AI-powered chatbot named Tay was released by Microsoft in 2016. The chatbot was designed to mimic the language and behavior of an 18-year-old American girl and was used to interact with people on Twitter. Unfortunately, within 24 hours of its release, Tay began to tweet offensive and racist comments which caused it to be taken down. This showed how AI can be easily manipulated and can lead to dangerous outcomes.

Who is accountable for such harm?

It is a matter of increasing concern that, when someone is harmed by Artificial Intelligence or a large language model or an app developed using that model, who should be accountable? It has been demonstrated that Artificial Intelligence can indeed cause injury. This leads to the question of who is at fault if the model produces an erroneous result or offensive or libelous material, and who should be liable for any resulting harm. It is unclear whether the model, the company that created it or the developer responsible for the app would be held responsible.

Whether Competition Laws would be affected?

The development and operation of certain models may require a hefty amount of computational resources, resulting in high costs that can limit access to smaller organizations or individuals. This could lead to a power imbalance among the few large companies and organizations that can afford to utilize them to their benefit. Unless we all have 200 million dollars to invest in building our own models, these monopolies may be unchallenged and lead to questions of whether competition and laws are being violated.

OpenAI's recent shift to a paid service further highlights the potential issues of models like PaLM by Google and CICERO by meta being owned and operated by a single entity. Additionally, it raises questions about the value people receive when data generated by them is used for these models.

Speech Regulation and Disparity

The generation of text on the internet is mainly in English and mainly from the western world, creating a disparity. This brings up difficult legal and ethical dilemmas, such as what kind of speech should be defended, what type should be regulated, and who should make these judgments.

Conclusion

In conclusion, the use of large language models can present a range of legal issues. These include intellectual property rights, copyright, privacy, data protection, and ethical considerations. It is important to be aware of these potential issues and be mindful of any applicable laws when using these models in any capacity. Taking the necessary steps to ensure compliance can help to mitigate any potential risks associated with the use of large language models.