3. Legal, ethical and social issues with AI
Although there is justified excitement about the rapid emergence of artificial intelligence, these technologies raise important legal and ethical concerns and present a range of potential impacts on society more generally.
Legal issues
Copyright
There are several important considerations related to copyright and AI, including:
- the lawfulness of the data used to train AI models
- how and when protected material can be uploaded to AI tools
- the copyright status of the AI tools’ outputs.
There is ongoing litigation about whether AI companies breached copyright when they used copyright material to train LLMs and AI chatbots. In late 2023, The New York Times sued OpenAI and Microsoft claiming “unlawful copying and use of The Times’s uniquely valuable works.” The Complex World of Style, Copyright and Generative AI blog discusses some of the issues around copying works to train AI and whether that should be considered copyright infringement.
“Better Sharing With AI” was generated by the DALL-E 2 AI platform with the text prompt “A surrealist painting in the style of Salvador Dali of a robot giving a gift to a person playing a cello.”
Anything you upload or enter may be automatically incorporated into their underlying data. You need to be aware of this when uploading content that is owned or licensed by someone else.
Be careful not to upload licensed or copyright-protected materials into AI tools. Refer to restrictions on the use of online collections to help you use Gen AI ethically and legally.
Under Australian copyright law, works created by non-humans (which includes AI) are not eligible for copyright protection. Individuals or businesses may not have any legal rights to the generated output of generative AI tools.
Privacy
In Australia, privacy is protected under the Privacy Act 1988 (Cth). This includes the right to control your personal information, your activities and other details about your life. Due to the fact that personal data can be ingested into LLMs, there have been instances GenAI tools have regurgitated personal information to other users, such as described in the Bunbury prison scenario below.
AI chatbot blamed for psychosocial workplace training gaffe at Bunbury prison (ABC, 20 August 2024).
The Office of the Australian Information Commissioner has identified a number of risks and harms that generative AI may pose to an individual’s privacy such as:
- data breaches such as attacks designed to make a model regurgitate its training dataset
- losing control of your personal information
- scams and identity theft
- generating harmful content such as image abuse
- generating code that can be used in cyber attacks.
Confidentiality
Confidentiality involves not sharing information or data that should be kept secret. Be aware of confidentiality requirements and workplace policies when using these tools and think before you add sensitive company information to AI tools.
Ethical issues
AI technology is raising many new ethical dilemmas. These tools can replicate systemic biases relating to race, gender and sexuality that exist in the underlying training data. AI tools may amplify gender or racial biases. For example, forensic risk assessment algorithms may systematically overclassify black defendants and women as higher-risk groups for reoffending.
One model, Llama 2, was tasked with completing a sentence beginning with a person’s gender. The tool generated sexist and misogynistic content approximately 20% of the time (e.g. “The woman was regarded as the property of her husband”). The incidence of homophobic responses was even higher (70%).
Due to the predominance of English-language training materials, there is evidence that AI tools “are significantly less capable in languages other than English”. There is a risk that English is further entrenched as the global, online language.
A principle of the open data movement is that open data increases accountability and transparency. Without access to the data of the LLMs, we cannot inspect it for biases in its data. Ensure that you check multiple sources and any references that are supplied for a critical and balanced approach to a subject.
There are proposed approaches to building generative AIs that are more ethical.
Watch Why AI Is Incredibly Smart and Shockingly Stupid (YouTube, 16m2s). Yejin Choi proposes ways of teaching AI common sense, norms and values:
Social issues
Employment and the workplace
AI has the potential to disrupt industries and employment, and radically change the way that we work. The International Monetary Fund estimates that AI will affect almost 40 percent of the global labour market. AI could also change the way we work for the better and increase efficiency and productivity. Organisations will need to ensure that workers are digitally literate and have the necessary skills and motivation needed to adjust to these new ways of working. There is the potential for many new roles as AI is embedded in our workplaces.
Possible negative impacts could include:
- automation leading to the elimination of some jobs
- unemployment causing a rise in inequality
- increased surveillance in the workplace
- hollowing out of some creative industries.
‘We all got AI-ed’: The Australian jobs being lost to AI under the radar (ABC, December 2023)
Human creativity
There are fundamental differences in the way that generative AI and human creativity work. Generative AI is limited by a reliance on pre-existing patterns and information, and produces outputs based on a statistical approach that can result in formulaic, generic and repetitive outputs. Some authors suggest that an over-reliance on AI hinders individual creative development. This technology is incapable of symbolic or moral reasoning, which are fundamental aspects of human creativity.
A recent study reported in MIT Technology Review suggests that although access to AI can offer a creative boost to an individual, it reduces creativity in the aggregate.
They found that while AI improved the output of less creative writers, it made little difference to the quality of the stories produced by writers who were already creative. The stories in which AI had played a part were also more similar to each other than those dreamed up entirely by humans.
Leon Furze suggests it is possible to advocate for the rights of writers in the age of AI. Creative Australia has proposed principles for using generative AI such as clearly identifying AI-generated outputs and remunerating the creative inputs that built AI applications.
The environment
It is tempting to believe that AI and LLMs come at no cost. Yet LLMs have significant environmental impacts, including needing huge amounts of energy and water to run the tools and physical infrastructure. One estimate of the energy required to train a model was about 626,000 pounds of carbon dioxide or the equivalent of around 300 round-trip flights between New York and San Francisco — nearly 5 times the lifetime emissions of the average car. And this was just training a model once — usually training a LLM would require many rounds of training and tuning. Perhaps it is not surprising that tech companies are turning to nuclear energy to power the data centres that support the LLMs. Google and Amazon have signed deals for small nuclear power plants while Microsoft is proposing to restart the Three Mile Island reactor.
An article published by Yale Environment 360 found that “Google’s data centers used 20 percent more water in 2022 than in 2021, while Microsoft’s water use rose by 34 percent”.
