Starmer is right to embrace AI – but it must be adopted ethically and carefully

Dr Keith Darlington

On Monday, Keir Starmer announced that the government will use Artificial Intelligence (AI) to transform the UK into a prosperous nation again. Britain already has a head start because it has been adopting AI through companies like DeepMind, and this should ensure it’s well-placed to benefit. Despite the lack of detail given by Starmer, I believe that the strategy is broadly right for the reasons I now offer.

In his speech, Starmer said that AI could unleash £47 billion in enhanced output in 3 to 4 years. That would be a massive fillip for the economy. Moreover, there are also countlessly many ways that AI can improve the lives of all of us – if adopted with care. AI applications come in all shapes and sizes: some are large-scale and apply to applications like robotics and creating smart infrastructures – such as roads. Whilst others may be small smartphone apps translating speech or language, or recognising an object in a photograph.

Welsh Benefits
Wales could benefit from this transformation because it will require the building of large data centres. – one of which is proposed to be built in Bridgend, South Wales. Plans are underway by the US technology company Vantage Data Centers to (https://www.constructionenquirer.com/2025/01/14/trio-of-tech-giants-unveil-14bn-ai-
data-centre-building-plans/) to build one of Europe’s largest data centre campuses in Wales. This will involve construction of 10-buildings and three substations at the former Ford car factory in Bridgend.

AI in Education
In education, AI is already being used to support teachers, for example, through assistance with lesson plans, generating test questions, or applying different learning methods. Students can use it to improve their understanding of core topics by using AI tools like ChatGPT. These systems are becoming more human-like in how they interact, and there are many more ways that could be used. There are endlessly many ways that AI can be used successfully in education.

AI in Healthcare
In healthcare, thanks to AI, patients can now get sophisticated scans analysed in minutes, enabling rapid diagnosis and treatment. Wales has already seen evidence of the value of AI implementation in diagnostic care in the NHS, with a digital pathology platform for prostate and breast cancer detection led by Betsi Cadwaladr seeing a 13% increase in prostate cancer detection rates (https://www.gov.wales/wales-forefront-uk-drive-turbocharge-ai).

One healthcare application of AI that has significantly helped me is in hearing. I was diagnosed with chronic tinnitus 27 years ago. Since then, I have had countless hearing tests in hospital ENT departments. Although necessary, these hearing tests are not something I have enjoyed because they involve sitting in a dark, soundproofed booth and responding to sounds created to measure your hearing with an audiologist.

This is time-consuming and costly for the NHS. However, this may soon be unnecessary because, in recent weeks, the computer company Apple announced air pods with built in hearing test features. They can complete a hearing test and deliver immediate results for those suffering from mild to moderate hearing loss. They can then work like hearing aids – as well as listening to music. As I have already found out, this can all be done in the comfort of your home in minutes. This is of benefit to the patient and reduces the demand for resources from the NHS.

Challenges
It has been well documented that AI automation could cause the loss of many jobs, but Starmer said little about this. AI automation will be disruptive, and politicians should be honest with us about this. Furthermore, Starmer also stated that the government will be giving AI tech companies access to anonymised NHS patient data will help them train new systems. There are strict rules in place around data confidentiality and privacy. However, the NHS has had problems with anonymising data in the past. Some think that health data can be made anonymous by removing the patient’s name from the patient record.

Unfortunately, it’s not that simple because data can come from multiple sources – some of which may be cross-references to reveal personal information. The biggest problem in these plans is likely to be the availability of energy to support the requirements of AI systems. The new AI, called generative AI, uses large language models that require copious amounts of power to train them. So much so that it has been estimated that AI could consume as much energy globally as the Netherlands by 2027 (https://www.bbc.co.uk/news/technology-67053139).

Many estimates suggest we won’t have enough energy to meet the energy demands of AI unless we rapidly adopt the use of small nuclear reactors. For example, the National Grid estimates that demand from AI data centres could be 6% of total consumption – up from 1% by 2030. Where is all this extra energy going to come from? It has often been said that AI will help with climate change, but not before these models have been trained and fully developed will we not see the net benefits in the short term. We must embrace AI, for we will all gain in the long run, but like all technologies, AI giveth and AI taketh. So, care must be taken to ensure our systems are safe, ethically appropriate, and offer good value for money.

Dr Keith Darlington is author of the book called Humanising AI – Bridging the Gap Between Artificial and Human Intelligence