Artificial intelligence (AI) has increasing been recognised and emphasized as an important priority for individuals, organizations, and governments. More than ever before academics, businesspeople, bureaucrats, and politicians alike have shown much interest in AI. As a matter of fact, many of the academics, businesspeople, bureaucrats and politicians believed that AI can play a critical role in improving performance, competitiveness and economic wealth.
The tremendous attention on Al is reflected in the dramatic increase in the number of books and research that addressed the importance of AI to organizations. Despite the significant increase in the amount of literature on AI over the years, the literature in this field has yet been able to provide a more comprehensive perspective of AI. The review of previous literature on AI reveals that limited attempts have been made to emphasize as well as explain issues related to its nature, advantages, and limitations.
Artificial intelligence as a technology enables technical systems to conduct activities that requires human intelligence such as learning, problem solving and decision-making. In the field of computer science, AI is used to develop machine learning (ML) models (as shown in Figure 1) that can learn to perform specific tasks. Over the years, as the AI technology progressed to a more advanced level, it was able to develop the improved foundation model (as illustrated in Figure 2) which replaced the task specific models. The foundation model can perform a range of general-purpose tasks.
The generative artificial intelligence (GAI) evolved from the AI and is powered by the foundation model [1]. Formally, the GAI can be defined as the machine learning model that generates new content in response to submitted user input (prompt/query). The GAI model generates content based on what it has learned from existing content obtained from the internet sources such as books, articles, and websites. The process of learning from the existing content is called training. The training process will result in the creation of a statistical model. The GAI used this statistical model to generate new content based on the given user input. The newly generated content can be in the form of images, text, audios, and videos.
Generative AI tools can enhance the customer experience for online travel agencies by generating personalized recommendations based on individual preferences. For instance, the Booking.com can recommend real-time travel updates for destinations, activities, accommodations based on the customer historical travel data [3]. The aviation businesses can also integrate chatbots to their websites to handle flight bookings, passengers’ inquiries, provide airline schedule as well as issuing alerts for schedule disruptions.
The generative AI can also help to optimize supply chain businesses [4]. As an example, the fuel consumption costs can be reduced by designing the efficient picking and delivery routes for drivers based on the traffic conditions and the priority of deliveries. The predictive maintenance can determine which machines or instruments that are most likely to fail. This allows manufacturers to schedule their maintenance efficiently as well as improve their overall equipment effectiveness (OEE).
Although it appears that as a technology, GAI is able to provide various advantages, it has several limitations related to the accuracy of its generated content. For instance, the generated content can be false, misleading, but presented as facts to the users. These misleading contents are known as hallucination [5]. For example, ChatGPT can generate fictional citations for an article that do not correspond to real sources. This hallucination occurred due to the insufficient training data as well as biases in data used to train the generative AI model. As such, it is crucial not to rely solely on the generated results before it is verified with the reliable sources to ensure factual accuracy.
While generative AI can provide solutions and complete tasks quickly, it cannot help in developing critical thinking skills. The overreliance of generative AI tools can also cause cognitive atrophy, which can be detriment to effective learning process of individuals in the long term. For example, students are often given sample articles to read and learn to develop their writing skills in classrooms. However, if they rely on the ChatGPT to complete their writing assignments [6] then they will not be able to develop their actual writing skills.
Equally important, generative AI also raises significant ethical concerns regarding the potential misuse or manipulation of the generated content that may unintentionally or intentionally defame individuals or organizations. For example, realistic videos (deepfakes) can be generated to make it appear as if a specific individual said or did something that he or she never did [7]. In addition, malicious content such as fake news or social media posts can also be generated to spread misinformation or propaganda that can be used to influence public opinion.
The training and utilization of generative AI model involves the use of powerful hardware as well as significant use of electrical energy. Recently, it has been found that the training and utilizing generative AI models can have negative environmental impacts on the planet in terms of their carbon dioxide emission. According to the study by [8], generating 1,000 images with Stable Diffusion XL can contribute to carbon dioxide emissions that is roughly equal to travelling 4.1 miles in a typical gasoline-powered car.
In summary, although generative AI has the potential to increase productivity, generate income and bring positive changes to the way we work, shop, communicate, it is also known that this technology can influence our learning and working environment negatively. While generative AI and its tools can enhance certain aspects of our works, they still lack general intelligence, creativity, and social awareness that humans possessed. In order to maximize benefits and mitigate risks of adopting the GAI, its adoption must be accompanied with proper skill development strategies, ethical governance, quality assurance mechanisms, as well as embedding it with values such as fairness and accountability.
References
- NVIDIA, “Generative AI – What is it and How Does it Work?,” NVIDIA. https://www.nvidia.com/en-us/glossary/generative-ai/
- Merritt, “What Are Foundation Models?,” NVIDIA, Jan. 24, 2024. https://blogs.nvidia.com/blog/what-are-foundation-models/
- L. Jockims, “ChatGPT and generative A.I. are already changing the way we book trips and travel,” CNBC, Apr. 22, 2023. https://www.cnbc.com/2023/04/22/how-chatgpt-generative-ai-are-changing-how-we-book-trips-and-travel.html
- Steinberg, “How supply chains benefit from using generative AI,” Ernst & Young, Jan. 9, 2024. https://www.ey.com/en_gl/insights/supply-chain/how-generative-ai-in-supply-chain-can-drive-value
- Cornell University, “Generative Artificial Intelligence,” Cornell University Centre for Teaching Innovation, 2024. https://teaching.cornell.edu/generative-artificial-intelligence
- Feijo and K. Ouellette, “What will the future of education look like in a world with generative AI,” MIT Open Learning, Dec. 18, 2023. https://openlearning.mit.edu/news/what-will-future-education-look-world-generative-ai
- Chan and A. Swenson, “Opinion: How to spot AI-generated deepfake images,” The Star, Mar. 21, 2024. https://www.thestar.com.my/tech/tech-news/2024/03/21/opinion-how-to-spot-ai-generated-deepfake-images
- S. Luccioni, Y. Jernite, and E. Strubell, “Power Hungry Processing: Watts Driving the Cost of AI Deployment?,” arXiv.org, Nov. 28, 2023. https://arxiv.org/abs/2311.16863