The Revolution of Chatbots

A term paper for “Media and Culture” in Spring 2023.

# 1    Introduction

Due to the desire to acquire Artificial General Intelligence (AGI), chatbots should be built in a way that acts as a tool, a toy, and a friend at the same time [1]. This realization came about when OpenAI released ChatGPT on November 30th, 2022. Although ChatGPT has not fully achieved the dream of AGI yet, it shows great potential.

A chatbot is defined as “a computer program designed to simulate conversation with human users” [1]. However, a more accurate definition for now would be “a chatbot is a representation of a Large Language Model (LLM) that integrates many other mechanisms, such as computer vision and robotics, while still attempting to achieve AGI.” It is an application of Artificial Intelligence (AI).

ChatGPT, which has remarkable capabilities such as text generation related to all known research fields, is a key milestone on the path to AGI. This attempt has sparked great curiosity and fears about the future world with AI. The purpose of this paper is to offer personal thoughts on chatbots and the future of a world with chatbots. In Section 2, we will take a glance at the current abilities of existing chatbots and the future plans of forward-thinking businesses. In Section 3, we will present other people’s concerns about AI and review the main techniques in LLM to draw conclusions about the future development of chatbots.

# 2    Abilities

Represented by ChatGPT and Midjourney, which is used in Discord, chatbots have demonstrated outstanding abilities in various fields related to digital text and images. In the first part of this section, we will focus on ChatGPT, whose killer feature is text generation. In the second part, we will delve into the state-of-the-art applications and realistic expectations of AI.

In the field of education, including mathematics and physics, ChatGPTs performance can reach or even exceed the average level. Additionally, ChatGPT has tremendous potential in text classification tasks, effectively addressing problems such as genre identification, sentiment analysis, stance detection, and more. In the areas of text generation and code generation, ChatGPT performs even better: it is difficult for people to distinguish ChatGPT’s responses from those of real humans. ChatGPT can be used for both simple and complex programming tasks. However, it’s important to note that in all fields, the chatbot might provide plausible-sounding but incorrect information [2]. As the first glimpse of AGI, when provided with accurate contexts, particularly in role-playing contexts, ChatGPT is able to respond like a friend, a teacher, or any other person, although there is no specific research specifying its performance in perceiving emotions. Overall, despite the existence of false responses, ChatGPT can offer not only human-like help in terms of information but also in terms of emotions.

Due to the fantastic abilities of ChatGPT, it is reasonable to conclude that some people could be replaced by ChatGPT [3]. Furthermore, AI has become an integral part of the future plans of global corporations. Among them, Microsoft has shown an extraordinary passion for AI. The global corporation has integrated many AI-powered co-pilots into its traditional products, such as Office 365 and Windows. Nvidia, a chip giant that provides high-performance graphics processing units (GPUs) for video games and AI applications, envisions a future with AI. In video games, players will encounter non-player characters (NPCs) that behave in a human-like manner, and the outcome of a single game may change depending on players’ interactions with NPCs. Overall, chatbots herald an imminent revolution in this era.

# 3    Concerns

The human-like performance of ChatGPT triggers concerns on AI. Some of them are from technics, while others are in the domain of ethos, which may even evolve into fears.

In terms of technics, people mostly worry about its development. Current chatbots faces certain limitations, such as insufficient understanding, energy consumption [2]. Additionally, since researchers have little knowledge about why LLMs show emergent abilities that are essential to the huge leap of ChatGPT at a certain computational scale, the scale of parameters, it is not fair to optimistically predict that in the future, the capabilities of LLMs will continuously increase rather than stay at this level [4]. This abrupt advance in LLM research arouses researchers’ interests [2]. However, it is not reasonable to say that researchers will soon figure out what happens in LLMs since the training data is a tremendous amount. Overall, the limited understanding currently, predictable huge energy consumption and unpredictable result-oriented LLMs result in people’s concerns in AI’s development.

In term of ethos, the concerns are even stronger. In May 31st in 2023, AI experts and public figures, including CEO of OpenAI, said that “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war. [5]” It seems that the risk of extinction brought by AI is closer than any other era. When it comes to security issues related to AI, a scenery of AI-powered robots killing humans may come up. However, by now, AI has not developed to that extent. The risks of extinction are mainly from the usage in cheating and breaking laws, because the data used in LLMs is invisible and the output of LLMs is unpredictable. The concerns focus on copyright, false information, malicious usage, bias and discrimination, privacy and data security [3]. Since AI is a new application of technology, the regulations and laws are not determined. For example, what if a chatbot shows an undisguised discrimination against a country, a race? Besides, there have been many applications of AI porn and cheating which might be malicious. Overall, the concerns of ethos are not fictional but realistic, which require policymakers and the public to enact laws to restrict the use of AI. Of course, it costs time.

# 4    Conclusion

This paper reviews the revolution of chatbots, highlighting their potential role in the development of AGI. It discusses the abilities demonstrated by chatbots, such as ChatGPT, and emphasizes the interest and investment in AI by major corporations that envision a future world driven by AI. However, concerns are raised regarding the development of chatbots, including technical and ethical considerations, which are expected to impede the progress of AI.

# References

  • [1] Adamopoulou, E.; Moussiades, L. An Overview of Chatbot Technology. Artif. Intell. Appl. Innov. 2020, 584, 373–383. [Google Scholar] [CrossRef]
  • [2] Yiheng Liu, Tianle Han, Siyuan Ma, Jiayue Zhang, Yuanyuan Yang, Jiaming Tian, Hao He, Antong Li, Mengshen He, Zhengliang Liu, et al. Summary of chatgpt/gpt-4 research and perspective towards the future of large language models. arXiv preprint arXiv:2304.01852, 2023.
  • [3] Tyna Eloundou, Sam Manning, Pamela Mishkin, Daniel Rock: GPTs are GPTs: An early look at the labor market impact potential of large language models. arXiv:2303.10130, 2023.
  • [4] Wei, J. et al. Emergent abilities of large language models. arXiv:2206.07682, 2022.
  • [5] AI experts and public figures: Statement on AI Risk. 2023. [Website]
« 日志(一)