As internet technology has grown and spread over the years, so too has the ways in which content within it is produced. In the past, users during the Web 1.0 era could only passively receive content. However, in Web 2.0 era, every user can participate in the creation of content on the internet. These are known as Professionally Generated Content (PGC) and User Generated Content (UGC), respectively. Aside from these two, some opinions suggest that artificial intelligence (AI) will play a more important role in content production in the future, referring to the approach known as Artificial Intelligence Generated Content (AIGC). An example of this is ChatGPT, an AI Chatbot created by AI research laboratory OpenAI, which has recently received extensive attention due to its high capabilities in generating content.
ChatGPT is a large-scale language model (LLM) 1, built on the foundation of GPT-3.5 and fine tuned 2 to further optimise it and upgrade its capabilities. OpenAI makes use of natural language processing technologies based on deep learning 3, alongside reinforcement learning 4 strategies in order to better train ChatGPT. Moreover, to further improve the performance of the program, OpenAI utilises human feedback in coordination with reinforcement learning [1] to train ChatGPT, adding an element of human opinion and decision making to achieve a more desirable and comprehensive level of language generation and understanding. As a result, through examining a large amount of currently existing text and dialogue, ChatGPT can communicate increasingly naturally and fluently. This is illustrated in the image below, where ChatGPT’s abilities in understanding human languages, as well as its proficiency in generating high-quality, bespoke responds can clearly be seen.
1 GPT-3 includes around 175 billion parameters and requires 800 GB of storage space. The amount of data that is used in its training has reached 45 TB, approximately 46080 GB.
2 Fine Tuning: A technique for transfer learning. The model designs and parameters are copied from the source model and fine tuned based on the requirements of the new model.
3 Deep Learning: A subset of Machine Learning that teaches computers to learn and process data in a way inspired by the human brain. It uses data or experience to automatically optimize artificial neural networks.
4 Reinforcement Learning: A branch of Machine Learning that rewards desired behaviours and punishes undesired ones, allowing the AI to perceive its environment and act in a way that maximises the benefits.
ChatGPT’s Popularity in The Business World
ChatGPT’s ability to generate human-like text in multiple different languages, its advanced performance and its versatility all contribute to making it an attractive tool in the business world. The American internet media, news and entertainment company BuzzFeed announced a partnership with OpenAI, stating that it would use ChatGPT in content creation as well as user personalisation. Influenced by this announcement, BuzzFeed’s stock price rose by over 300% over a period 2 days. Firms and companies from varying fields, such as the e-commerce platform Secoo and the enterprise AI software provider C3.ai have followed suit, each announcing plans to integrate ChatGPT into their own products. Hongyi Zhou, CEO of internet security company Qihoo 360, summarised this notion in the interview column ‘Dialogue Underneath the Stars’: “in three to five years, various industries will be reshaped by GPT, so whoever does not embrace artificial intelligence now will be eliminated”.
ChatGPT’s Business Applications: Marketing, Customer Service, Cost Control
So, what potential applications does ChatGPT have in the business world? In terms of business marketing, ChatGPT can rapidly create large amounts of high-quality promotional pieces, including articles, blogs, social media posts as well as descriptions of the product, all based on the individual interests of the user as well as requirements of the business. This allows companies to create content that better fits their target audience. Additionally, ChatGPT provides marketing teams with a greater range of ideas and possibilities, as well as giving teams more time to focus on planning and analysis, improving operational efficiency. Furthermore, by increasing the number of promotional pieces on social media, companies can increase exposure for their goods and services and obtain high quality traffic.
Moreover, in terms of customer service, ChatGPT can be integrated into customer service chatbots in order to provide customers with a more natural, human-like and personalised conversation experience. Chatbots that have ChatGPT integrated can efficiently handle a wide variety of commonly seen customer requests, lowering the need for human interference and allowing the customer service team to focus their time and efforts on more complex or higher value tasks. Aside from that, because of its real time, fast response speed and its ability to process multiple segments of dialogue, ChatGPT is able to provide quick, around-the-clock customer service, thereby increasing the satisfaction of customers, decreasing the possibilities for customer complaints and demands for refunds, thereby decreasing the costs required to resolve such disputes.
The application of ChatGPT within the commercial world is a topic that has massive potential, though much of it still remains unexplored. However, these applications must be tailored to suit each individual company, keeping in mind their requirements and regulations. As such, factors such as industry type, legal considerations and ethical obligations must all be considered. Despite this, the future development of AI programs such as ChatGPT has the potential to greatly benefit businesses, helping them to grow and better serve their consumers. If carefully planned and implemented, ChatGPT has the power to change how companies conduct business, enhancing them and driving industry efficiency, innovation, and success.
Reference:
[1] Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C. L., Mishkin, P., … & Lowe, R. (2022). Training language models to follow instructions with human feedback. arXiv preprint arXiv:2203.02155.
Author：Ziquan OU (City University of Hong Kong)
The work described in this article was supported by InnoHK initiative, The Government of the HKSAR, and Laboratory for AI-Powered Financial Technologies.
(AIFT strives but cannot guarantee the accuracy and reliability of the content, and will not be responsible for any loss or damage caused by any inaccuracy or omission.)
