This article is written by Mehrdad Safaei, a Data and Artificial Intelligence Program Analyst at Digital Academy, Canada School of Public Service (CSPS), Government of Canada – the Digital Academy group involved in developing data and AI courses for public servants of Canada.

They are also a research assistant at Simon Fraser University and Johnson Shoyama Graduate School of Public Policy (JSGS), University of Regina. Their recent researches include tweeter sentiment analysis for ‘Attributions of Blame and Credit in Policy-Making’ and using NLP algorithms to generate briefing notes ‘The end of the policy analyst? Investigating the capability of artificial intelligence to generate plausible, persuasive, and useful policy analysis’.

The views expressed in this paper are solely those of the author and do not necessarily reflect the views of the Canada School of Public Service (CSPS), nor those of the Government of Canada.

  • The problem: Large generative AI models (LGAIMs) such as ChatGPT and Stable Diffusion are transforming industry at pace, but they bring challenges as well as benefits.
  • Why it matters: We need to ensure that LGAIMs are used responsibly and ethically to minimise their potential harms.
  • The solution: Develop effective strategies to address challenges to new generative AI technologies.

Image of AI toolsPicture 1: Artificial Intelligence Tools / DALL.E

As artificial intelligence (AI) continues to evolve, large generative AI models (LGAIMs) like ChatGPT and Stable Diffusion are transforming the way we work and communicate. However, with great power comes great responsibility. In this blog post, we'll take a closer look at the opportunities and challenges of using ChatGPT in various industries, including banking, hospitality, tourism and information technology.

One of the most significant benefits of ChatGPT is its ability to enhance business activities, such as management and marketing. However, there are also potential consequences to consider, such as biases, disruptions to practices and threats to privacy and security. To address these concerns, further research is needed. This includes identifying the skills and resources required to handle generative AI, examining biases attributable to training datasets, exploring contexts best suited for generative AI implementation, determining the optimal balance of human and generative AI for various tasks, assessing the accuracy of text produced by generative AI and uncovering ethical and legal issues in using generative AI across different contexts.

At the heart of this discussion is the need for responsible use of LGAIMs. As we explore the potential of these powerful tools, we must also consider their impact on society and individuals. By developing effective strategies for addressing these challenges, we can ensure that LGAIMs are used responsibly and ethically, maximising their potential benefits and minimising their potential harms.

Large generative AI models (LGAIMs)

Large generative AI models (LGAIMs) are a type of unsupervised or semi-supervised machine learning algorithm that uses existing content like text, images, audio, video, and even code to create new content that resembles human language, images, or sound. Their main purpose is to generate original content that seems real and human-authored, such as sophisticated text. LGAIMs like GPT-4 (Generative Pre-trained Transformer 4) or Stable Diffusion are trained on massive datasets and can produce highly sophisticated outputs. Their potential applications are numerous, including chatbots, natural language processing and content generation.

Current LGAIMs such as GPT-4 and Stable Diffusion are some of the most advanced AI models available today. GPT-4 is a language generation model created by OpenAI that can produce highly sophisticated text that is often indistinguishable from text written by a human. It has been trained on an enormous amount of data from the internet and can perform a wide range of language tasks, including translation, summarisation and question-answering. On the other hand, Stable Diffusion is a generative model that can produce high-quality images and videos. It is based on a diffusion process that allows it to generate complex images with realistic textures and shapes. Both GPT-4 and Stable Diffusion are examples of LGAIMs that have the potential to transform the way we communicate, create and innovate. However, they also present significant ethical and legal challenges, and it is crucial to carefully consider their implications for society and individuals.

Opportunities of using LGAIMs

Applications in banking, hospitality, tourism, and information technology industries

The integration of technology in banking is not a new phenomenon, as banks have always adopted new technologies to transform their operations. With the emergence of ChatGPT, there is an opportunity for banks to explore its potential applications in marketing financial services, data analysis and enhancing customer experiences. However, given the high regulation of the banking sector, technology adoption is often strategically explored. ChatGPT can be used for content creation, emotional appeals and personalised customer offers. While it presents opportunities for banks, limitations and potential consequences need to be considered. Trust in service provision and the impact on vulnerable consumers are some of the concerns that need to be addressed. Banks would be expected to invest in infrastructure and explore technical capabilities and human resources to integrate ChatGPT into their existing digital transformation strategies.

The use of ChatGPT can help both tourists and organisations by providing quick and accurate information, creating trip itineraries, recommending travel services and activities and generating marketing content. ChatGPT can also assist with back-office functions and enhance value-added services, such as menu engineering and recipe development. With multilingual support, ChatGPT can improve the overall travel experience and increase satisfaction. The use of ChatGPT in these industries can revolutionise customer communication, improve service, streamline operations and access knowledge databanks.

According to a recent study[1] , artificial intelligence, particularly the generative nature of ChatGPT, can contribute new ideas and concepts and help human team members better understand their problem and solution space. It has the potential to become a member of a hybrid innovation team by acting as an innovator in the new product development process. Additionally, recent reports have shown that AI can play a role in software development, assisting with code writing, automating simple tasks and error management in various phases of development.

Challenges of using LGAIMs

Bias, security, reliability and privacy

Recent studies[2] have highlighted several limitations and potential consequences of LGAIMs, including biases, disruptions to practices and threats to privacy and security. LGAIMs can perpetuate biases present in their training data, leading to discriminatory outcomes for certain groups of people. Furthermore, the use of LGAIMs can disrupt established practices, such as decision-making processes in industries like finance and healthcare. There are also concerns about the security and privacy of personal data that LGAIMs collect and process. If these systems are compromised, it could have significant consequences for individuals and organisations alike. Therefore, it is essential to address these limitations and potential consequences of LGAIMs to ensure that they are developed and used in an ethical and responsible manner.

Preparing for the future of work

A new Goldman Sachs report suggests that as many as two-thirds of jobs in the US and Europe could be automated to some extent. The report notes that while significant disruption to the labour market is likely, historically, technological progress not only makes jobs redundant but also creates new ones. The use of AI technology could boost labour productivity growth and boost global GDP by up to 7% over time. Certain jobs, such as those requiring physical work, are less likely to be significantly affected. In the US, office and administrative support jobs have the highest proportion of tasks that could be automated, followed by legal work and tasks within architecture and engineering. Countries such as Hong Kong, Israel, Japan, Sweden and the US are expected to be the most affected by AI automation. However, it is not yet clear how disruptive AI will be in reality.

Need for a pause

Tech industry executives and academics, including Elon Musk and Steve Wozniak, have signed an open letter calling for a six-month pause on large-scale experiments with artificial intelligence. The letter warns that companies researching AI are in an “out-of-control race” to develop and deploy more powerful digital minds that no one can understand, predict or control. It calls for governments to intervene if a pause cannot be quickly enacted, to avoid potentially apocalyptic scenarios such as the development of non-human minds that could eventually outsmart and replace us. While some ethicists criticise the letter for focusing on theoretical harms, industry watchdogs warn that companies are testing new AI technology on the public without considering broader consequences.

Call for further research

As technology rapidly evolves, it is crucial to explore and research various topics to strike a balance between AI and human capabilities[3]. In order to achieve this balance, further research is necessary in the following areas:

  • Identifying the necessary skills, resources and capabilities needed to effectively handle generative AI.
  • Examining the potential biases that can arise from training datasets and processes and how to mitigate them.
  • Exploring the most suitable business and societal contexts for implementing generative AI.
  • Determining the optimal combinations of human and generative AI for various tasks.
  • Developing ways to assess the accuracy and reliability of text produced by generative AI.
  • Investigating ethical and legal considerations surrounding the use of generative AI in different contexts.

By addressing these research topics, we can better understand the potential benefits and drawbacks of integrating generative AI into various industries and contexts, while ensuring that it is implemented in a responsible and ethical manner.

👋 You can write an article like this one! Share your thoughts with a community of public servants. Learn more

(Image Credit: Unsplash)