Chat GPT Friend of Foe? Chat GPT Friend of Foe? Ankit Dahal University of Northampton, UK Naya Aayam Multi-Disciplinary Institute (NAMI) Kathmandu, Nepal ankit.2021123@nami.edu.np
Abstract: Background: ChatGPT, a powerful language model developed by OpenAI, has made significant advances in natural language processing, leading to various applications in chatbots, language processing and other areas. However, the mode’s unprecedented ability to generate coherent and human-like responses has raised concerns about its potential misuse in spreading misinformation, cyberbullying, or even promoting hate speech. Results: This technical report analyzes a comprehensive review of the literature and case studies to assess ChatGPT’s impact on human society. The findings suggest that while ChatGPT has enormous potential, its use must be approached with caution. It is vulnerable to various forms of bias and tends to hallucinate between answers. Its ability to generate highly coherent and human-like responses raises concerns about its potential misuse. Conclusion: This technical report proposes that it is necessary for developers and policymakers to address the potential hazards of AI use by implementing strategies to prevent misuse, and ensuring transparency and accountability throughout the development and deployment process. This paper adds to the current discourse on responsible AI usage and its effects on society. Keywords: ChatGPT, Natural language processing, Deep learning, Neural network, Ethics, Responsible AI, Misinformation, Regulation, Security, Social implications, Text generation
- Introduction A chatbot is a type of computer program that emulates human conversation via text or speech. These conversational agents utilize natural language processing techniques to interpret and generate responses to user inputs. They serve various purposes such as customer service, sales, and personal assistance. Chatbots have gained significant popularity in recent years due to their ability to efficiently handle repetitive and straightforward tasks while providing round-the-clock service to users. Early history of Chatbots The history of chatbots dates to the mid-1960s when Joseph Weizenbaum created ELIZA, a program that could simulate a conversation between a therapist and a patient. Weizenabum’s work on ELIZA was published in the Communications of the ACM Journal in 1966 (Wallace, 2009). In the initial version of the chatbot called DOCTOR, ELIZA played the part of a Rogerian psychotherapist, using open-ended questions to engage the user in conversation. By doing so, she directed the focus away from herself and towards the user. To the surprise of the developers, people began to attribute human-like qualities to ELIZA and started to trust her with their personal stories, sensitive information, and confidential secrets (Neff, 2016). The other prominent early chatbots were PARRY and Racter, that were developed in the 1970s and 1980s. PARRY, also known as "ELIZA with attitude," was developed by Kenneth Colby in the mid-1970s. Unlike ELIZA, which was designed to simulate a conversation between a therapist and a patient, PARRY was programmed to mimic the behavior of a person with paranoid schizophrenia. PARRY used natural language processing techniques to respond to user inputs and was designed to respond in a paranoid and defensive manner. Colby's work on PARRY was published in the journal "Artificial Intelligence" in 1976 (Colby et al., 1976). Racter developed by William Chamberlain and Thomas Etter was designed to generate text using a technique known as "grammar transformation," in which sentences were broken down into their grammatical components and then reassembled in a different order to create new sentences. Chamberlain and Etter's work on Racter was published in the book "The Policeman's Beard is Half Constructed" in 1984 (Neff, 2016). Current landscape of Chatbots In recent year, chatbots have become more sophisticated and widespread, thanks to major advances in NLP and machine learning. Today, chatbots are used in a variety of industries, from e-commerce to healthcare, and are becoming increasingly popular as a way for businesses to improve customer service and engagement. Chatbots can improve customer engagement, increase efficiency and productivity, save costs, operate 24/7, and improve customer satisfaction (Dwivedi et al., 2021; Noreen et al., 2023; Selamat and Windasari, 2021; Zhang et al., 2023). With the advent of smartphones, people gained access to a plethora of information through the internet. In 2010, Apple Inc. developed SIRI, their own virtual assistant that was based on chatbot concepts and could be used on smartphones and other devices. This development was well-received, especially since the internet was widely accessible. Other tech giants like Google, Microsoft, and Amazon followed suit, creating their own virtual assistants, such as GOOGLE NOW, CORTANA, and ALEXA, respectively, between 2012 and 2015. The rise of Artificial Intelligence technology led to more people using social media to conduct their business online. Additionally, internet access became more affordable and widespread, thereby fueling the rapid growth of Artificial Intelligence
Types of Chatbots There are two primary types of chatbots: rule-based and machine learning-based. Rule-based chatbots rely on pre-defined rules and decision trees to direct their interactions with users and are simple and easy to implement, but they may not be effective at handling complex or personalized customer inquiries. On the other hand, machine learning-based chatbots learn and adapt their responses over time based on previous interactions and data. These chatbots can handle a broader range of user inputs and offer more personalized responses, but they require more advanced technical expertise and resources to develop and maintain. The utilization of chatbots for customer engagement and support is becoming increasingly popular among businesses. According to Adewole (2020) and Jabarooti & Tan (2019), rule-based chatbots are one type of chatbot that utilize pre-defined rules to provide scripted responses to customer inquiries. In contrast, AI-based chatbots, which employ natural language processing and machine learning algorithms to offer personalized responses to customer inquiries, are another type (“International Journal of Machine Learning,” n.d.; Solanki et al., 2019). Moreover, chatbots can be categorized based on their intended use case, such as customer service, sales, or marketing (Liu X, 2021). They can also be classified based on their deployment, such as on a website, messaging app, or social media platform (Raza, 2019).
ChatGPT