Could Generative AI and ChatGPT improve the delivery of Public Services ?
Generative AI and ChatGPT have become very popular over the past few months, offering new opportunities and potential benefits for the delivery of services.
Generative AI is a form of artificial intelligence (AI) that is used to generate new content from existing data. ChatGPT is a natural language processing (NLP) system that can generate automated conversations with humans, mimicking the way people interact with each other.
Here , we will explore the impact of generative AI and ChatGPT on the delivery of public services, considering the benefits, risks, and opportunities they present.
The Benefits
The most obvious benefit of generative AI and ChatGPT is their ability to automate mundane tasks such as customer service interactions, reducing the need for human resources for some (or many) repetitive or simple interactions. For example, a chat-bot could answer frequently asked questions from customers or provide basic information about services offered by a government agency. By automating simple tasks, governments can reduce costs associated with hiring personnel or outsourced services, to answer enquiries.
For more complex queries, it can be used to analyse large amounts of data quickly and accurately without needing additional human resources for analysis. This could help government make better decisions based on data-driven insights.
The combination of advanced Natural Language Processing capability, the ability to learn, to simplify and then summarise large quantities of day, will also provide the potential to improve access to public services. For instance, it could be used to interact with citizens 24/7 (both passively and actively), delivering information about available services, warnings or allow them to apply for certain services online without having to wait in line at the office.
This could reduce waiting times and enable governments to provide better service to citizens who need it most.
The advanced NLP capability combined with learning ability can help drive a truly user centric design, with the creation of personalised experiences for each and every citizen by tailoring content and advice to their individual needs and preferences. Consider a conversation, where the service can determine the requirements of the enquirer, any special needs, and combine the understanding to deliver a personalised response.
Public Sector Health
Generative AI has the potential to revolutionise medical processes by providing a more efficient and accurate diagnosis process (reviewing and summarising symptoms, results, records etc). It could automate many of the onerous tasks associated with medical diagnosis, including data gathering and analysis, as well as identifying patterns in medical data that may lead to a diagnosis. It could also be applied to the huge datasets of NHS data that is gathered everyday, to understand trends, spot opportunities and automatically prioritise service delivery.
And before we jump into that Utopian (but unlikely) world where your GP has been replaced by an Alexa app, there are more immediate opportunities that could provide real benefit. A Generate AI solution that can understand what the doctor, nurse, GP is doing, and saying, and automatically capture all of the key information needed to maintain records (sufficiently detailed and understood to provide an authoritative record). Imagine; no more need to manually fill out forms, record patient interactions, key observations, diagnosis or treatment — the Generative AI would have not only recorded the patient interactions, but structured the notes, saved them, validated it, and if requested, provide additional guidance or information that the medical staff hasn’t considered.
Some of the Risks
As wonderful and incredible this next step in AI is demonstrating to be, there are also risks associated with this technology, especially relating to the delivery of public services.
One major risk is trust: many citizens will be wary of interacting with an automated system instead of a real person when seeking advice or applying for services. Moreover, privacy concerns will need to be considered when collecting personal data from citizens in order to customise content or make decisions based on algorithmic insights. There is also the risk that AI systems could be biased if they are not properly trained, or utilise datasets that do not represent all members of society equally.
Generative AI is not an Expert System
Generative AI and ChatGPT are not expert systems, but they can learn from their research and analysis to refine their answers. Stephen Wolfram’s recent article provides an excellent example of what can and can’t be done, and offers insight into how these risks could be tackled.
ChatGPT (and the other systems out there) works through training on a large amounts of text, (such as Wikipedia articles or other sources), and then uses an algorithm to generate responses. The AI can then refine its responses based on its research, analysis and feed back it receives.
Hence, Generative AI may not be able to distinguish between common myths and known facts. It could provide incorrect information or lead people down a path of misinformation if it does not understand the context of the query correctly. It also relies heavily on the quality of the data used for training, which could lead to responses that are not appropriate for the question being asked. (For example, a thousand ‘Bots’ arguing “Black is White, and Red is Green” may lead to a confused response until corrected).
Additionally, since Generative AI is not an expert system, it may not be able to answer complex questions correctly without human intervention (again, Stephen Wolfram’s article gives a simple but compelling answer to this).
In order to mitigate these risks, Generative AI needs be used in conjunction with other tools such as fact-checking algorithms, Knowledge Systems and human curation of content, to help ensure that the information provided is accurate and up-to-date. Generative AI will need to be used alongside human experts in order to verify complex queries, ensure quality, provide more detailed answers when necessary, and resolve misinterpretations where the underlying data isn’t sufficient to provide clarity or accuracy.
By combining the strengths of both methods (Generative AI and Knowledge/Expert Systems), it will be possible to create more reliable responses while still taking advantage of the speed and scalability of AI technology.
Back to our exploration in the Public medical environment, there is a risk that generative AI might miss important aspects of a patient’s condition or fail to identify potential risks associated with certain treatments or diagnoses. How will this be treated and managed from the aspect of the citizen? These ethical, legal and patient dimensions will need to be understood and tackled, for any service to develop, grow and survive.
Policy, Governance and Transparency
In order to fully realise the potential benefits of Generative AI while minimising the risks involved in their use in public service delivery, Agencies and Authorities will need to take steps to ensure they are implemented and managed responsibly:-
- making sure that any data collected from citizens is stored securely;
- that any automated decisions made by algorithms are explainable; and
- that any technology deployed is designed with user experience in mind so as not to create an impersonal interaction between its citizens and the government
Summary
Governments will need to invest in developing ethical frameworks for using these technologies responsibly so as to ensure there isn’t unfair advantages for certain groups, that doesn’t discriminate, and responsibly identifies and handles errors in decision-making.
Overall, Generative AI and ChatGPT offer many potential benefits for improving public service delivery but also come with certain risks that must be managed responsibly. With these risks managed, proper implementation strategies in place and the investment in the ethical frameworks for responsible use, governments could unlock an array of opportunities to make large improvements to quality, efficiency and availability of services.
In a world of shrinking funding, increases in citizen demand, and urgency to improve, Generate AI provides an exciting and tantalizing opportunity that will require clear structure and understanding in order to meet expectations and Return on Investment.
References
1. Wolfram|Alpha as the Way to Bring Computational Knowledge Superpowers to ChatGPT
For more Blogs and Posts on ‘Best Practices’, ‘Security’ and Architecture : https://www.cloud-dog.com/blogs/
Twitter: https://twitter.com/garyseymour
