Paper “Large Language Models: Their Success and Impact” appeared in Forecasting

Authors: Spyros Makridakis, Fotios Petropoulos, Yanfei Kang*

Journal: Forecasting

DOI

Summary

ChatGPT, a state-of-the-art large language model (LLM), is revolutionizing the AI field by exhibiting humanlike skills in a range of tasks that include understanding and answering natural language questions, translating languages, writing code, passing professional exams, and even composing poetry, among its other abilities. ChatGPT has gained an immense popularity since its launch, amassing 100 million active monthly users in just two months, thereby establishing itself as the fastest-growing consumer application to date. This paper discusses the reasons for its success as well as the future prospects of similar large language models (LLMs), with an emphasis on their potential impact on forecasting, a specialized and domain-specific field. This is achieved by first comparing the correctness of the answers of the standard ChatGPT and a custom one, trained using published papers from a subfield of forecasting where the answers to the questions asked are known, allowing us to determine their correctness compared to those of the two ChatGPT versions. Then, we also compare the responses of the two versions on how judgmental adjustments to the statistical/ML forecasts should be applied by firms to improve their accuracy. The paper concludes by considering the future of LLMs and their impact on all aspects of our life and work, as well as on the field of forecasting specifically. Finally, the conclusion section is generated by ChatGPT, which was provided with a condensed version of this paper and asked to write a four-paragraph conclusion.

The article discusses the success of ChatGPT, a large language model (LLM) that has reached 100 million active monthly users in just two months. The success of ChatGPT is attributed to its ability to communicate in natural language, its willingness to answer questions, and its free-of-charge nature. However, there are criticisms associated with its use, such as job loss and perpetuating biases, as well as technical problems such as “hallucinations” where it provides responses that sound convincing but have no basis in reality.

The paper compares the accuracy of two GPTs (ChatGPT and CustomGPT) when posed with questions regarding the M forecasting competitions. CustomGPT was trained using published papers from a subfield of forecasting only where the answers to the questions asked are known. The results showed that CustomGPT was able to provide more accurate and helpful responses than ChatGPT in most cases. However, neither GPT was perfect and caution should be used when using information provided by language models blindly.

The article discusses the use of chatbots to improve the accuracy of judgmental forecasts. Two chatbots, ChatGPT and CustomGPT, were used to answer questions related to how to adjust statistical/ML forecasts and how to improve management meetings that judgmentally adjust statistical/ML forecasts. CustomGPT provided more specific advice than ChatGPT.

Finally, the article discusses the future of natural language processing (NLP) models such as ChatGPT and their impact on society. It is noted that NLP models have been adopted by many organizations around the world and more firms are considering its adoption to improve their efficiency and add value. The current capabilities of ChatGPT are just a stepping stone for its future potentials as AI technology is rapidly advancing due to intense competition.

Finally, to end this conclusion with a human touch, we present two contradictory views about what ChatGPT can do and its future. According to the first, ChatGPT will “probably remain just a tool that does inefficient work more efficiently” with nothing to worry about [15]. In the second view, we must get prepared for the coming AI storm [16]. This is an old concern for new technologies. The Luddites, for instance, broke machines because they believed that new technologies would lead to massive unemployment and negatively affect their jobs. We now know they were wrong and the new technologies increased rather than decreased employment by creating extra jobs. LLMs will not be an exception. Even so, it may take some time until their advantages are fully exploited and their disadvantages minimized. It is part of human nature to overreact to the potentially threatening LLM technology, but time has shown repeatedly that humans have a great ability to adapt to difficult situations by turning problems into opportunities, and ChatGPT will provide the opportunity to further advance technological progress and improve the quality of life on Earth. Our end objective would be to implement a vertical LLM specifically trained for forecasting tasks, a “ForecastGPT”. This LLM should be trained on the entirety of the forecasting literature, as opposed to specific tasks that we presented on this paper, with an aim to offer informed and complete responses to all forecasting knowledge.