This article first appeared in Digital Edge, The Edge Malaysia Weekly on December 25, 2023 - January 7, 2024

In 2023, there was scarcely a conversation that did not touch on generative artificial intelligence (AI) platforms such as ChatGPT. But amid the hype were dire warnings from experts about the risks of unrestrained AI development.

Could 2024 be the year in which the use and development of AI will be tamped down and regulated? Digital Edge turned to ChatGPT — the latest version, GPT-4 — for its thoughts. 

The conversation has been edited for brevity.

ChatGPT: Predicting the exact future of generative AI for 2024 would involve speculation, but there are several trends and potential advancements that could shape its trajectory based on the developments up to my last update in April 2023:

•     Increased accessibility, with platforms enabling individuals with little to no technical background to create content, whether that be images, text or code;

•     Improved quality and fidelity, including better handling of nuances in human communication;

•     Expansion into new domains such as game development, interactive entertainment, real-time voice translation and advanced simulation for research purposes;

•     Ethical and legal frameworks around privacy implications, ethical use, intellectual property rights and potential biases. This might result in new regulations and standards for generative AI applications;

•     Integration with other technologies such as virtual reality and augmented reality;

•     Economic impact, where disruption in certain industries could lead to changes in job markets;

•     Collaborative AI that works alongside humans;

•     Personalised experiences, such as custom education plans and individualised entertainment;

•     AI ethics and creativity, with discussions around the nature of creativity and whether AI can truly be creative or if it is simply a tool used by humans to extend their own creativity;

•     Advanced natural language understanding; and

•     Generative AI in scientific research.

Here are some key elements that are often considered important in AI regulations:

•     AI systems should be transparent. Users should be informed when they are interacting with an AI and have access to information about how decisions are made within the system;

•     Clear lines of accountability should be established. This involves ensuring there is always a human ultimately responsible for decisions made with the assistance of AI systems;

•     AI regulations should enforce strict data protection standards to ensure personal information is not misused;

•     AI systems should be designed and regularly tested to ensure they do not perpetuate biases or discrimination;

•     AI should be aligned with ethical standards, which includes not only respecting human rights but also considering broader impacts on society and the environment;

•     AI systems should be reliable and function as intended, with errors and failures kept to a minimum to avoid harmful impacts;

•     There should be a requirement for human oversight, especially in critical areas such as law enforcement, judicial decisions and other areas where AI decisions can have significant consequences;

•     Organisations may be required to conduct impact assessments before deploying AI systems, especially in high-stakes scenarios, to understand and mitigate potential risks;

•     Certain uses of AI may be restricted, such as facial recognition in public spaces or decision-making systems without adequate transparency; and

•     Encourage or require training for professionals who develop or work with AI, to ensure they understand the ethical implications and responsibilities.

ChatGPT: Here are some strategies that can be employed:

•     Improve the training process of AI models. This includes training on larger, more diverse and high-quality datasets that are regularly updated with factual information;

•     Implementing automated fact-checking tools within generative AI platforms;

•     Educate users on the limitations of AI and promote digital literacy;

•     Establish procedures for human oversight where content generated is reviewed;

•     Enforce transparency measures that require AI-generated content to be clearly labelled;

•     Develop and use digital watermarking and provenance tools to trace AI-generated content back to its source;

•     Create regulatory frameworks that hold platforms accountable for dissemination of false information;

•     Encourage development of ethical guidelines and standards for AI use;

•     Implement measures to prevent use of generative AI for malicious purposes;

•     Platforms can engage their communities in governance, allowing users to report misinformation; and

•     Develop AI that can recognise its own uncertainty and indicate when it is generating content outside its knowledge base or when its confidence level is low.

ChatGPT: Whether this is fair or not can be contextualised through several lenses:

•     National security entities like the Space Force handle classified and sensitive information. Generative AI platforms, especially those hosted by third parties, could potentially expose such information. The ban can be seen as a measure to prevent data leaks and espionage;

•     The integrity of operations in a military branch is paramount. If generative AI is deemed a risk that could compromise this integrity, restricting its use is a reasonable step;

•     Given the novelty of generative AI and its potential risks, a ban can be a temporary, preventative measure while a comprehensive assessment and establishment of secure usage policies take place; and

•     AI systems may sometimes generate outputs that are not entirely accurate or reliable. In a military context, reliance on such unverified information could have serious consequences.

On the other hand, the ban could be argued against on several grounds:

•     Restricting the use of generative AI could hinder the exploration of potentially beneficial applications that could advance the Space Force’s capabilities;

•     If proper cybersecurity measures and data handling protocols are in place, an outright ban might be seen as overly cautious and a missed opportunity for operational improvement;

•     Other nations or entities that leverage AI more effectively could gain a strategic advantage over the US Space Force; and

•     Generative AI can significantly improve efficiency and productivity. Banning it might limit employees’ ability to perform certain tasks effectively.

Ultimately, the fairness of the ban depends on the balance between the need to protect sensitive information and the potential benefits of generative AI. A fair approach might involve a nuanced policy that allows for controlled use of generative AI with strict compliance and security protocols, rather than an outright ban.

Yes, as a general rule, it’s wise to avoid sharing sensitive personal information when interacting with AI services like ChatGPT.

Even though data used for model training is typically anonymised and stripped of identifying details to protect privacy, it’s still prudent to err on the side of caution. This is particularly important because once data is incorporated into a training dataset, it’s difficult to extract or delete individual entries.

ChatGPT: To my knowledge, as [at] April 2023, there have been no such legal actions taken by Jodi Picoult or George R R Martin against OpenAI for copyright infringement.

ChatGPT: (Browses the web)

Indeed, there has been a lawsuit filed by The Authors Guild and a group of authors against OpenAI. According to the claims, the lawsuit was filed in the Southern District of New York for copyright infringement, specifically concerning the use of their works of fiction in training the GPT models.

ChatGPT: The lawsuit filed by The Authors Guild and prominent writers against OpenAI raises complex legal questions about copyright and the use of copyrighted materials in training AI models like ChatGPT. The determination of guilt or liability in this case would depend on the legal process and is ultimately for a court to decide.

The plaintiffs, which include authors such as Jodi Picoult and George R R Martin, allege that OpenAI used their copyrighted works without authorisation to train its AI models, which they claim constitutes copyright infringement.

OpenAI’s defence and the outcome of the case would likely hinge on interpretations of fair use, the transformative nature of AI training processes, and the specifics of copyright law as it applies to machine learning.

The legal proceedings would consider various factors, such as the nature of the AI’s training data, the way the AI generates content, and the impact on the market for the original works. The issue is precedent-setting and could have significant implications for the future of AI development and the rights of content creators.

It’s important to note that ‘guilt’ in civil litigation, such as copyright infringement cases, is typically referred to in terms of ‘liability’ rather than the criminal law concept of guilt. The lawsuit’s proceedings will provide a clearer understanding of these legal boundaries as the case evolves.

Save by subscribing to us for your print and/or digital copy.

P/S: The Edge is also available on Apple's App Store and Android's Google Play.