In recent years, Artificial Intelligence (AI) has come out of Science Fiction and into the realm of the real, especially since Open AI launched ChatGPT into the public consciousness.
In the meantime, Dr Gopichand Katragadda, IET President for 2023 – 2024, took the view he would use his time in office and his expertise in the field, ‘…To sharpen the IET’s focus on Digital Futures and Sustainability.’
To get to the core of the issue, the IET organised a series of webinars to discuss three aspects of AI:
- How can we co-create an IET Digital Futures position?
- How can we promote equality, diversity, and inclusion?
- How can we strengthen the IET’s global impact on AI?
These discussions were held in three separate events, each with three webinar sessions. They’re meant to highlight how industry and society can best direct and use Artificial Intelligence (AI) efficiently and ethically - in order that smarter decisions can be taken more quickly when and where they are needed.
Some insights
How can we co-create an IET Digital Futures position?
May 1, 2024
The discussion series, and the report, makes the point that science and engineering work within a continuum of technological development. What is possible now was unthinkable 30 years ago. In that time computational power has grown enormously, as has the explosion of growth of the internet, its user base and the data that it generates. In turn this data and other technological developments have helped fuel the rapid growth in the capability of AI.
Just as using the internet became an essential part of engineering, so too will AI become an essential skill for engineers. Engineers will always require knowledge of engineering fundamentals, the underpinning theory, and the engineering mindset. Even so, relying on this expertise alone will no longer be enough, engineers will need to become digitalists alongside their existing skillset.
One of the upsides of AI’s data crunching potential is the space this frees up for engineers to be more creative, use critical thinking skills and take decisions.
There are two critical points here. One is that AI has a remarkable capacity to sort ‘the gems from the garbage.’ It can whittle down a database of 1.5 million research records to nine relevant records in seconds. The second point is that working out the accuracy of this process depends on the model used to carry out the task, as AI is a tool of augmentation, not a replacement for engineers. At best, AI is no more than 98% accurate (meaning 2% inaccurate), something that could prove critical. The imprecision of AI means any outputs require checking and validation by humans (i.e. engineers).
How can we promote equality, diversity, and inclusion?
June 6, 2024
Data bias is a difficulty in creating an AI, which can lead to incorrect decisions. Rigour is central to eliminating bias from the data used to train AI models. As a result, 80% of the time spent developing an AI is spent annotating it and training networks so that everything under the data ‘umbrella’ is fair.
Some techniques can achieve this. Involving diverse groups of people can mitigate data biases, allowing AI to recognise people from different ethnicities. Introducing people with different perspectives into the design process of AI can create fresh, more innovative, and creative approaches and methods in data modelling. AI will become more relevant to the people who will use it soon.
Using languages beyond English and other European languages would make the technology more inclusive. While, taking account of cultural sensitivities and differences would make AI more useful and accurate. It would also mean AI being accessible and beneficial to all demographics.
It is a paradox that as our technology becomes more sophisticated, the less accessible it becomes to the elderly, or people with sight, hearing, cognition, or memory impairments. Similarly, 50% of people on the planet are still affected by a digital divide.
Data can become more accurate, and decisions more focused on reality by having diverse people working within the design and data processing process. Another factor might be introducing more diverse data sets into the design, testing and training process.
How can we strengthen the IET’s global impact on AI?
July 2, 2024
The EU’s recent AI Act highlights the importance of where data is, whether it is protected and whether it represents a broad spectrum of people. Indeed, it is now possible to interact with AI through the language of your choice, a sign that AI is democratising?
One risk of AI is that its use in schools could create a situation where the AI does the work, not the pupils. A balance between the use of AI and educational settings needs to be struck. Parents and teachers would both welcome more information to familiarise them with the risks and benefits of AI.
A recurring concern in these discussions is data bias, something that creeps in because it has been collected and input by people. To develop a system truly representative, the database must span the broad spectrum of the human condition. From neurodiversity to employment, ethnicity, age, and disability. One danger of human input is that as data is updated, it can become increasingly subject to bias.
AI data will never be 100% accurate, even with the necessary checks and balances, reliability, and accountability guardrails we deem necessary. It can be productive in, for example, a healthcare setting, where it can assist the physician in any decisions they make. Because the healthcare specialist is responsible for any decisions taken in a clinical setting, the AI can never be more than an assistant.
Another area of concern is the vast amounts of energy consumed by data centres and the importance of building sustainability ethics into the sector. This would help reduce energy consumption and the production of greenhouse gas emissions.
Gathering feedback from experts
The webinars themselves were unique in that they were not simply discussions between the panellists with audience questions messaged to the panellists. These discussions were open for other participants to take part. This created a less exclusive, more collaborative event that could go beyond a narrow, received opinion.
Some recommendations
| Topic Areas | Recommendation | Who It Concerns |
| Understanding AI’s opportunities and global impact. Promote ethical AI development. |
Establish AI ethics committees to develop and promote comprehensive ethical guidelines and standards for AI. The focus should be on maintaining high standards for data quality and AI system transparency. |
IET, Industry, Government, Academia |
|
AI for equality and inclusion. Bias mitigation |
Ensure that AI systems are trained on diverse and representative datasets. Regularly audit algorithms for biases and use fairness metrics to ensure equitable outcomes. | Industry, Academia, Government |
|
Develop AI solutions to address societal challenges. Healthcare innovations |
Support the development of AI applications in healthcare. Examples include predictive analytics for disease outbreaks, personalised medicine, and telehealth services. | IET, Industry, Government, Academia |
|
AI risk management Invest in emerging AI applications. |
Focus on sectors where AI can have a significant impact, like healthcare, automotive, agriculture, and entertainment. Innovations in these areas include adaptive learning, legal process automation, and health and wellness applications. | Industry, Academia, IET |
For a fuller understanding of the discussion that occurred across the series of webinars, the themes, and concerns, along with a set of recommendations, a breakdown of the panellists and a foreword from Dr Gopichand Katragadda himself, why not check out our report:
Leading with intelligence: AI’s role in digital futures and sustainability
----------------------------------------------------------
What Do You Think?
What do you think we can be doing to achieve the goals for the Digital Future and Sustainability with and for AI?