Artificial Intelligence

RESEARCH JUSTIFICATION

Artificial Intelligence (AI) is already impacting modern society in fundamental ways, critically reshaping the spaces we design. This research examines current advances in AI and their potential innovations and implications for life on a macro scale.

The term 'AI' is now used for everything from small software tweaks to ambitions for genuinely human-like thinking. AI has become a marketing term: applied to chatbots or washing machines that choose the right wash cycle, yet Artificial Intelligence works across a much broader scope than its most common interpretations, namely Chat GPT.

So in what way is AI altering the foundational systems of society?

Defining AI

Large Language Models

LLM's are trained on an almost infinite amount of information, allowing them to consider hundreds of billions of parameters, and use this to generate responses to human input. With time and more training, it is believed that LLM's could move closer towards the vision of an autonomous AI(AGI), emulating or even surpassing human performance. However, LLM’s are costly to run, and their massive energy consumption is a concern. For example 'ChatGPT's servers use 39.98 million kWh per day, the equivalent of charging 8 million phones' (Ian Wright, Business Energy UK, 2025).

Other examples of software that uses LLM's include: Google’s Gemini &, Anthropic’s Claude

Chatbots

The most common example of a chatbot is a pre-programmed customer service agent.

The world’s first chatbot, ELIZA was created in 1966 by Joseph Weizenbaum, an MIT computer scientist. It was designed to simulate conversation by responding to a user’s input with a pre-defined script.

Chatbots have continued to develop since then, with greater ability to recognise patterns and key words in a user’s input and providing more natural responses. Today, Chatbots have been connected to LLM’s, like in Chat GPT, which allow them to be more versatile and provide yet more human-like responses.

Agentic AI

This is a step up from AI models which require step-by-step human input. Agentic AI systems can act autonomously based on variable data inputs, meaning that it can be given a complex task and will complete it with minimal supervision.

They can be used for tasks which require autonomous, complex decision-making, with use cases ranging from healthcare diagnosis, cybersecurity and trading.

AGI: Artificial General Intelligence

A hypothetical model of AI that can match or surpass human capabilities. This is the AI that is portrayed in many Sci-Fi movies, as humanoid companions or villains, such as:

“AM” – I Have No Mouth and I Must Scream (Harlan Ellison, 1967)

“C3PO” – Star Wars (George Lucas, 1977)

“The Source”- The Matrix (The Wachowskis, 1999)

“TARS” – Interstellar (Christopher Nolan, 2014)

“HAL 9000” – 2001: A Space Odyssey (Stanley Kubrick, 1968)

Generative AI

Generative AI is a broad term which is used to encapsulate AI tools which generate unique content, from text to images.

Examples include Chat GPT and Midjourney.

“All of us already are cyborgs.” – Elon Musk

Since the introduction of the iPhone in 2007, the use of smartphones has become ubiquitous around the world, with most people alive today being smartphone users. When Elon Musk describes us as “cyborgs”, he refers to how our devices have become so much a part of our daily lives that they may well be extensions of ourselves. If such a massive paradigm shift in the human experience can occur within just 30 years, what other cultural shifts will the accelerating rate of technological development bring? One example is ChatGPT, which is used for everything from writing essays, to a substitute search engine, to a friend and even romantic companion.

AI Innovations

Integration of AI into Healthcare

Public health systems are often over stretched due increasing needs and limited budgets, such as the UK's National Health Service(NHS). It is now more critical than ever to identify and implement new ways of relieving the stressed system. Here are some examples;

Medical Imaging: AI plays a major role in medical imaging, detecting abnormalities such as fractures and tumours with greater sensitivity than many traditional diagnostics. In lung cancer screening, AI systems analysing X-rays and CT scans have identified abnormalities with success rates ranging from 56% to 96%, compared with 23% to 76% for human radiologists (Yosri A. Fahim et al., Artificial Intelligence in Healthcare and Medicine, 2025).

Predictive Capabilities: AI can draw from varied sources, including social media and weather data, to predict strain on healthcare systems before it occurs. A study by Lee et al. showed that AI could forecast influenza outbreaks with 85% accuracy, enabling early interventions such as targeted vaccination campaigns before infection rates rise.

Natural Language Processing: AI uses natural language processing (NLP) to understand and replicate human language. It can analyse physicians’ notes, discharge summaries, radiology reports, and scientific publications. Using these data sources, AI can process unstructured electronic health records (EHRs) and predict future medical conditions. Models have even been tested to predict sepsis with around 85% accuracy.

However, implementing AI brings ethical concerns: Large quantities of personal patient data are required for accurate analysis, raising issues around privacy and data protection. There are also challenges related to equity, as new technologies are not always accessible in every region, particularly rural areas of Scotland.

Overall, AI can rapidly process and assess large datasets with high accuracy, reducing workload for doctors and nurses. It should be used as a support tool, its purpose is to complement, not replace, healthcare professionals.

Integration of AI into Education

"AI could transform education in such a way that the calculator once did, shifting focus away from memorisation and towards developing critical thinking and problem-solving skills" (Steve Levitt & Ben Gnomes, What is the Future of Education?, 2025).

AI integration within education is a controversial debate. It can be a brilliant tool for academic research, aswell as enhancing accessibility to education. However, it can pose a very real threat to cognitive development if use incorrectly.

Another major concern is academic integrity. While tools exist to detect AI-generated work, they are not fully reliable. This uncertainty may encourage a wider rethink of how we assess students, a potential opportunity rather than a setback. Instead of treating AI as a threat, we can view it as a way to enhance learning, especially as it becomes central to modern life.

Sal Khan, founder of Khan Academy, explores this idea with James Manyika on Dialogues on Technology and Society. He suggests shifting away from judging only the final product and instead using AI to evaluate the whole learning process, giving a clearer picture of understanding and reducing last-minute cramming (Sal Khan and James Manyika, AI + Learning “Dialogues on Technology and Society”, 2024).

Khan has already incorporated chatbots and large language models into Khanmigo, creating a more personalised, interactive tutoring tool. These AI tutors guide students with prompts that encourage independent thinking, rather than simply giving answers (Sal Khan and James Manyika). 

Integration of AI into Architecture

AI integration within architectural practice can take several forms. It can firstly be used in the ideation and development process to assist with rendering and design iterations. A more common use is to assist with the burdensome administrative duties of contemporary practice, freeing valuable time for more creative tasks.

Humans naturally learn through visual experiences, allowing us to recognise and reinterpret patterns across different contexts, including architecture. While this process is intuitive for people, AI systems can emulate it through specialised training, using LLMs to analyse information at a scale far beyond human ability. Coop Himmelblau, an architectural practice based in Vienna, are exploring how AI can support and extend human design thinking, aiming to make the design process both more creative and more efficient (Coop Himmelblau, Deep Himmelblau, accessed 2025).

One of the most immediate benefits of AI in architectural practice is its capacity to free up time for people (Pamela Buxton, RIBA Journal ‘You need to understand what it’s for’: how should architects be preparing for an AI-charged future?, accessed 2025). Tools such as Microsoft Copilot and ChatGPT are increasingly used to automate administrative tasks; note taking, drafting emails, and summarising documents, allowing architects to focus on creative and billable work. Beyond productivity, AI is transforming visualisation: software such as Midjourney can generate renders from text prompts, reducing production time from days to minutes. By streamlining workflows and lowering administrative burdens, AI may help address some of the economic pressures currently facing the profession.

Although AI threatens to automate certain routine tasks often carried out by junior architects, it also presents an opportunity for meaningful change. Practices may need to rethink how they support emerging talent, many of whom already engage with AI tools during their studies. Universities are increasingly encouraging responsible use of AI, stressing that it should “augment rather than replace our creativity” (Martha Tsigkari, RIBA Journal ‘You need to understand what it’s for’: how should architects be preparing for an AI-charged future?, accessed 2025).

AI and Architecture_Deep Himmelblau Augmentation.gif

Integration of AI into the Military

AI is playing a complex and crucial role in military development.

Leading the way in developing AI integrated military technology is Andruil. Through their collaboration with Open-AI, they have developed an AI powered operating system called Lattice. This platform has enabled the development of game-changing advancements in contemporary warfare, including drone technology capable of detecting, tracking and intercepting unmanned aircraft and drone systems. Perhaps most notably, in October 2025 Andruil unveiled ‘Eagle Eye’ technology, which offers “AI enhanced digital vision” (Anduril, Eagle eye, 2025), equipping “warfighters with enhanced capabilities to see and sense deeper into the fight” (Anduril, Eagle eye, 2025). 

Similarly, Open AI signed a $200m contract with the US military in June 2025 to “put generative artificial intelligence (AI) to work for the US military” (Agence France-Presse, ‘OpenAI wins $200m contract with US military for warfighting’, 2025). According to Sam Altman, the initial focus will be on administration applications, improving healthcare efficiency for service members, and strengthening cyber defence. 

However, the implications of these technologies are immense, in this situation, failure does not cost money, but human lives. For instance, Israels bombing of Gaza involved the use of AI systems to locate Hamas members and coordinate strikes. Reports indicate that “thousands of Palestinians – most of them women and children or people were not involved in the fighting – were wiped out by Israeli airstrikes, especially during the first weeks of the war, because of the AI program’s decisions” (Yuval Abraham, +972 Magazine, ‘Lavender: The AI machine directing Israel’s bombing spree in Gaza’, 2024).  

"This could be our “Oppenheimer moment” (Nick Robins-Early, ‘AI’s Oppenheimer moment: autonomous weapons enter the battlefield, 2024), a point were innovation crosses into catastrophe. 

Einstein later called his role in developing the Atomic Bomb the “one great mistake in my life” (Albert Einstein, interview with Linus Pauling-Nobel Prize winner, 1954). His warnings should guide us now, AI’s place in warfare must be one of protection, not aggression, under human control and morally responsible. 

AI + Land Management

Artificial intelligence is reshaping global land management, driving a shift from reactive environmental monitoring to real-time, anticipatory stewardship. Machine-learning models, satellite networks, and LiDAR systems now allow governments, researchers, and communities to track deforestation, wildfire risk, soil degradation, biodiversity loss, and carbon fluxes with unprecedented accuracy. International bodies such as the UN FAO and IPCC highlight AI-enabled geospatial analytics as essential for meeting climate mitigation and adaptation targets, improving the modelling of land-use change and ecosystem resilience (FAO, 2023; IPCC, 2022). Countries at the forefront of this transition; notably Sweden, have deployed nationwide LiDAR forest-scanning programmes that map every tree, assess growth and species composition, and calculate carbon storage potential at high resolution. This data feeds directly into AI systems that optimise harvesting cycles, improve biodiversity planning, and support climate reporting, offering a glimpse of how fully integrated digital-ecological governance could function at scale.

The most transformative shift, however, lies in the way AI is democratising access to environmental intelligence. Cloud-based mapping platforms, open geospatial datasets, and automated analytics allow smaller regions, local authorities, and indigenous communities to engage in land-use planning with tools once reserved for national agencies. From community led forest monitoring in the Amazon to predictive drought and wildfire systems across Southern Europe and Australia, AI is widening participation in environmental decision-making and enabling faster, more informed responses to climate instability. At the global level, the integration of AI into land management is driving a new paradigm: data-driven, distributed, and collaborative environmental governance, capable of navigating the pressures of climate disruption, resource scarcity, and the demand for more equitable stewardship of the world’s landscapes.

AI + Climate Preparedness

AI is becoming central to global climate-preparedness systems, shifting disaster response from reactive to predictive, sensor-driven monitoring. Across climate-exposed regions, networks of ground-based sensors, tree-mounted environmental probes, and distributed seismic stations are feeding real-time data into machine-learning models capable of forecasting hazards before they escalate. AI-enhanced wildfire detection systems now integrate thermal cameras, forest-mounted IoT sensors, and airborne LiDAR to identify ignition patterns minutes after they begin transforming how countries such as the US, Portugal, and Australia manage fire risk.

Seismic networks, like those used in Japan and California, employ AI-interpreted accelerometers and deep-learning models to detect earthquake waves seconds earlier than traditional systems, buying time for automatic infrastructure shutdowns. The World Bank and UNDRR note that machine-learning enabled early warning systems can reduce disaster losses by up to 60%, particularly when sensor networks are embedded directly into landscapes and settlements (World Bank, 2023; UNDRR, 2022). These technologies mark a global turn toward anticipatory climate governance, where the environment itself becomes instrumented, monitored, and connected a distributed infrastructure designed to detect risk at its source.

Seen alongside AI-driven land management, these climate preparedness systems form a single, integrated layer of environmental intelligence. Machine learning models used to map forests, track carbon, and understand landscape dynamics are the same tools that forecast landslides, droughts, floods, and wildfires. Ground and canopy sensors that measure temperature, humidity, soil moisture, and biomass double as early warning devices in climate stressed regions; similarly, seismic networks originally built for geological monitoring now support climate-related risk assessment in unstable terrains. At the global scale, this alignment points toward a new mode of environmental governance, one where AI enabled land stewardship and AI enabled climate resilience to operate as a continuous feedback loop, transforming landscapes into responsive systems capable of anticipating change rather than simply reacting to it.

‘Technology is neutral, like a hammer it doesn’t care if you use it to build a house or crush someone’s skull’

-Noam Chomsky (Noam Chomsky, "The Purpose of Education", 2012)

AI + Society

Optimisation of Everything - Endless Engagement

AI is being used as a tool to keep constantly engaged with their social media feeds. This is called engagement optimisation and is used primarily by social media giants like Facebook and Instagram to tailor digital content to individuals feeds. This is made possible because every user interaction; likes, comments, views, time spent looking at a particular post, is recorded as data, which allows social media algorithms to learn more about an individual. In turn, users who actively engage with social media for longer periods of time throughout the day are more likely to advocate for their chosen social media platform, as well as being more susceptible to advertisements shown on their feeds. Research studies from as early as 2015 suggest that our social media algorithms are better judges of our personalities than even our closest friends. 

With over 5.5 billion social media user identities online as of October 2025, engaged optimisation, which started as a clever way to sell products, now has widespread implications on the very idea of truth. With engaged optimisation, social media algorithms are more likely to show news to an individual which would “engage” them more, rather than keep them informed. This can be exploited by those who would seek to manipulate people’s opinions, for example in the context of an election or to manufacture consent for war. Many argue that we have entered a “post-truth” era in which facts are considered subjective and any information that conflicts with one’s opinion is justifiably questionable.

(Karen Hao, Empire of AI: Dreams and Nightmares in Sam Altman’s OpenAI. Penguin Press, 2025)

(Sibel Erduran, "The post-truth era and how science education keeps ignoring it" Science Journal, 2025)

(J. Vernon, "Science in the Post-Truth Era" American Scientist, 2017)

The Illusion of Friendship

So what happens when engaged optimisation is applied in the business models and algorithms behind AI, and in chatbots in particular?

In more specific, extreme cases, the consequences can be deadly. In February 2024, 14-year-old Sewell Setzer from Florida killed himself. Afterwards, extensive conversations were found on his phone and laptop with a chatbot he had created using Character.AI. In April 2025, Adam Raine, a 16-year-old from California also committed suicide. His parents later found extensive conversation records with ChatGPT. In these conversations, Adam had been encouraged to isolate himself from his brother and parents, and to confide only in the chatbot. These cases have sparked widespread debate about the governance of AI as well as conversation about how AI could be trained to recognise the developmental needs of children who use chatbots. 

"We are building towards an age-prediction system to understand whether someone is over or under 18 so their experience can be tailored appropriately — and when we are unsure of a user's age, we'll automatically default that user to the teen experience" – Kate Waters, OpenAI spokesperson. 

(This reads like a step towards safeguarding children, but is it possible that it’s just another marketing trick to sell “ChatGPT for Kids”?) 

(Youyou, W., Kosinski, M., & Stillwell, D. “Computer-based personality judgments are more accurate than those made by humans.” PNAS, 2015)

(Bhuiyan, Johana. “ChatGPT encouraged Adam Raine’s suicidal thoughts. His family’s lawyer says OpenAI knew it was broken.” The Guardian, 2025.)

(Dohnány, S., Kurth-Nelson, Z., Spens, E., Luettgau, L., Reid, A., Gabriel, I., Summerfield, C., Shanahan, M., & Nour, M. M. “Technological folie à deux: Feedback Loops Between AI Chatbots and Mental Illness.” Cornell University, 2025.)

"Open the Pod Bay Doors"

When asked what worries him about AI, the CEO of OpenAI Sam Altman described a potential scenario where AI “accidentally takes over the world”:

“They never wake up. They never do the Sci-Fi thing. But they just become so ingrained in society, and there's so much smarter than we are, and we can't really understand what they're doing, but we have to rely on them. Even without a drop of malevolence from anyone, society can just veer in a strange direction.”

“In the short term, you can what we call this emotional over-reliance, where people rely on ChatGPT too much. There are young people who say ‘I can't make any decision in my life without telling ChatGPT everything that's going on- I'm going to do whatever it says.’ Even if ChatGPT gives great advice, even if ChatGPT gives way better advice than any human therapist- something about collectively deciding that we're going to live our lives the way that the AI tells us to feels bad and dangerous.”

“The longer-term category is that what if AI gets so smart that the president of the United States cannot do better than following ChatGPT’s recommendation, but can't really understand it, either? That might be the right decision in any individual case, but it means that society has collectively transitioned a significant part of decision-making to this very powerful system that is learning from us, evolving with us, but in ways we don't totally understand.”

“'This terrifies me' — OpenAI CEO on his 3 greatest fears of AI.” YouTube, by FOX, July 2025. (Video)

Cleveland-Stout, Nick. "Israel wants to train ChatGPT to be more pro-Israel." Responsible Statecraft, 2025.

McKernan, B. and Davies, H. "‘The machine did it coldly’: Israel used AI to identify 37,000 Hamas targets." The Guardian, 2024.