The societal value of AI: The need for ethics and the path forward

‘It is completely possible to innovate with AI and benefit from these technologies in ethical and safe ways’

In my role as chief data scientist and head of Responsible AI at UST, I advise large UK and global companies on how they can best benefit from AI, along with helping them to achieve these benefits.

Since November 2022, when OpenAI launched ChatGPT, the explosion of interest in AI cannot really be understated. The intrigue in a chatbot that could write stories and computer code quickly spread across society, AI went mainstream.

Since then, we’ve seen the emergence of a race amongst the large technology companies such as Google, Meta, Microsoft and OpenAI, each competing with the other to have the smartest large language model (LLM), the kind of machine learning model that powers most forms of Generative AI (GenAI).

And recently the geopolitics of AI have come to the fore, as China’s DeepSeek shows the rest of the world how GenAI that competes with OpenAI’s leading-edge models can be developed at a fraction of the cost, even if DeepSeek’s LLM builds on top of Meta’s open-source Llama model, using methods developed by OpenAI. Less than two weeks before DeepSeek hit the news, along with revoking Biden’s AI Executive Order (essentially the US’ plans for AI regulation), president Trump announced a $500 billion investment in the US’ Stargate, a programme designed to ensure the US stays ahead in what is becoming the global AI arms race.

All this points to a need for governments to plan how they position and secure themselves in this race. For example, the UK Government is considering the development of a Sovereign LLM, in part to mitigate the risk of over-reliance on overseas-built AI in an increasingly uncertain global environment.

If this rapid pace of development coupled with the geopolitics carries risk, to some degree it also causes a loss of focus on the reasons why we’re building AI.

So, why are we building AI?

That is not meant to be a naïve question. AI can be a force for good and there are many examples of this. Research has shown it improves medical diagnostics, and so globally has the capacity to reduce the burden on increasingly pressured health services. In conjunction with satellite images AI is being used to spot forest fires before they become uncontrollable. Enterprises are using and further developing AI to assist them with their sustainability goals.

Along with keeping a focus on the good AI can achieve within society, it’s also important to consider the need to ensure it’s used ethically: managing bias, hallucination and the unintended consequences that can occur when we use AI.

AI is built on data that records information about society. We live in a society that is unfortunately inherently biased, and the data collected from and by society often reflects these biases. AI can reflect and magnify these biases and so needs to be managed to ensure it is fair.

The importance of ethics

When an LLM model generates content, it does so by predicting sequences of events. These predictions can be wrong. To put it simply, the LLM predicts the incorrect word or line of code and continues to do so through a sequence or chain of events. When this happens, the content created can appear highly relevant (in the sense that it is very well written or logically structured and relevant code) despite being completely incorrect. This has become known as AI hallucination.

Sometimes the consequences of implementing AI is not as is intended. For example, in Spring 2024, Google’s Vertex model generated images of Nazis who were of diverse ethnicities, leading to an outcry.

Google was seeking to ensure equality and diversity in its GenAI, but in doing so actually enabled it to generate such offensive images. While this is a high-profile example of AI not working as intended, there will likely be many more across society as AI is used more widely, unless consideration to this is given at the time of AI design and implementation.

This really matters in areas such as health and education, when for example it’s important to ensure care pathways aided by AI do not discriminate against certain patient groups and school pupils graded by AI are graded fairly across all demographics.

As the pace of AI development accelerates and AI becomes increasingly geopolitical, it will likely prove challenging to consistently keep a focus on AI as a force for good and on ensuring it is used ethically. But there are practical steps businesses and other organisations can take to manage the risks of AI error and unintended consequences. It is completely possible to innovate with AI and benefit from these technologies in ethical and safe ways. What’s required right now is a wider acknowledgement and understanding of why the ethical use of AI is important.

Throughout my career I’ve innovated with AI while always seeking to ensure it’s being used ethically. Together AI innovation and ethics can ensure we achieve good with AI, while seeking to minimise the unintended consequences.

Heather Dawe is chief data scientist and head of responsible AI at UST