The ChatGPT Conundrum

Chat Generative Pre-trained Transformer, or ChatGPT as it’s commonly called, has ushered in a new era of human-computer interaction. With its ability to read and understand user input, and generate answers accordingly, ChatGPT has garnered monumental fame in a short span of time. Opinions about it however, are rather divided. While some believe that ChatGPT is the next big thing that will revolutionise the way we work, others feel that it will bring about unprecedented security and privacy concerns. The solution, however, might lie somewhere in between.

In just a few months of its release, ChatGPT has gained exponential global popularity and has made inroads into the workplace. Employees now use it for one or more of the following:

1. Text Prompts: to generate and improve prose and code development, summarise text, classify content, answer questions and translate language.?
2. Prompt Engineering without Application Program Interfaces (APIs): as a manual workflow or automated via methods such as Robotic Process Automation (RPA) in conjunction with other technologies.?
3. Prompt Engineering with APIs: This process allows you to directly integrate ChatGPT with a broad spectrum of applications using APIs.
4. Automation: automate complex or redundant tasks in order to make room for project related tasks, thereby increasing productivity.
5. Business Planning: organise work and create strategies to achieve business goals in a time bound manner.

ChatGPT can clearly unlock some great value in organisations. However, many individuals and organisations have serious reservations around its use, few of them include:

1. Jobs Takeover: fear of the loss of livelihood to AI. Infact, according to Telsyte’s Australian Digital Consumer Study 2023, 35 percent of workers anticipate their roles could be performed fully by machines, AI, or robots in the future.?
2. Privacy: There are reservations about feeding ChatGPT with sensitive data that might be stolen by unauthorised third parties. This is why companies like PwC have restricted its use only for personal work, while banning its use for client tasks. Other companies have only allowed its use to restricted teams working in controlled IT environments, but have blocked its wider use by staff.
3. Bias: Ultimately, ChatGPT is a reinforcement machine learning (ML) model and any ML model is only as good as the data it’s trained on. Thus, there might be the probability of it overfitting a particular dataset and developing a bias.?
4. Environment: Training powerful ML models requires a high amount of computational power, which in turn requires a high amount of electricity, coming at the cost of burning fossil fuels. A recent research found that just in training GPT-3 alone, Microsoft, which is partnered with OpenAI, consumed a whopping 185,000 gallons of water to cool their data processing centres — which is, per their calculations, equivalent to the amount of water needed to cool a nuclear reactor!

When the pros of ChatGPT are considered against the cons, one may see that the necessity to neither completely rely on ChatGPT as a source of truth nor outrightly ban its use altogether. Instead, the solution lies in a balance between both, which may be achieved through some of the following ways:

1. Have a Policy: It is essential that organisations have internal conversations and deliberations on the use of ChatGPT for work and develop a definite policy on its use. Embracing the technology this way might give the organisation a competitive advantage.?
2. Bridge Responsible AI Gap: Many organisations have a policy but are not able to translate those high level principles to tangible actions in incorporating AI. This is called the Responsible AI gap, and can be bridged by:

• Empowering responsible AI leadership
• Establishing human and AI governance
• Conducting reviews for responsible AI
• Integrating tools and methods
• Building and testing a response plan

1. Mitigate Risks: This can be done by one or more of the following ways:

• Classifying what qualifies as sensitive data
• Encouraging the use of ChatGPT to aid rather than replace existing positions. This can also be achieved by upcoming AI such as Microsoft’s Copilot.?
• Learning to prompt with adequate information so that the bias is minimised and accurate results are obtained
• Use free tools such as BCG’s codecarbon to reduce CO2 emissions from your computing and FACET to achieve human-explainable AI.

At JOST&Co, we can help your organisation design policies around the use of ChatGPT and other AI, and help translate them into changes that enhance productivity. Contact us at 1300 791 950? - we’d love to hear from you.

References:

• Australian Financial Review ‘AI is meant to improve your job” (https://bit.ly/3oRPyzW)
• Australian Financial Review ‘PwC warns against using ChatGPT for client work’ (https://bit.ly/3oIDAbD)
• The Byte ‘ChatGPT is consuming a staggering amount of water’ (https://bit.ly/3L4RMDa)
• Gartner ‘Your 7 biggest questions about ChatGPT answered’ (https://bit.ly/446YVM9)
• Boston Consulting Group ‘Deliver Powerful Business Results with Responsible AI’ (https://bit.ly/3HlbOs2)