AI Safety, Ethics & Our Relationship with Technology

Technology should serve humanity

Artificial intelligence presents us with extraordinary opportunities. It can help us solve complex problems, improve the way we work, accelerate scientific discovery and expand what human beings can achieve.

But technological progress alone does not guarantee progress for humanity.

As intelligent systems become more capable, we need to consider not only what they can do, but how they should be designed, governed and used. We need to understand their limitations, anticipate unintended consequences and develop safeguards that allow innovation to flourish without losing sight of human interests.

My interest in AI safety, ethics and technology comes from a broader conviction: intelligence is most useful when it operates within systems designed to make good outcomes more likely.

I approach these questions as a systems thinker, drawing on experience in technology, change management, banking controls and the practical challenges of making complex organisations work.

The aim is not to slow technological progress for its own sake. It is to make that progress more sustainable, responsible and beneficial.

Why AI safety and ethics matter

AI safety is often discussed in terms of extreme future scenarios. Those questions deserve attention, but responsible development also requires us to examine the systems being deployed today.

How do we establish accountability when an automated decision causes harm? How do we distinguish reliable information from plausible-sounding misinformation? How do we prevent small weaknesses in individual components from becoming significant failures across a larger system?

These are questions of architecture, governance, incentives and control.

An intelligent system does not operate in isolation. It exists within a network of developers, users, organisations, data sources, infrastructure and regulatory requirements. Its effects depend on the relationships between those elements as much as on the capabilities of the model itself.

That is why I believe effective AI governance needs to look beyond individual models and examine the wider environment in which they operate.

From principles to practical safeguards

Ethical principles are important, but principles alone do not guarantee ethical outcomes.

Organisations need practical mechanisms for translating their intentions into repeatable processes. These can include clearly defined responsibilities, access controls, monitoring, independent evaluation, escalation procedures and documented accountability.

My work explores how systems thinking can help bridge the gap between abstract principles and operational safeguards.

One area of this work is PACER — Persistent Architecture for Control and Ethical Regulation of Intelligent Systems. PACER explores a framework for integrating control, ethical regulation and oversight into intelligent systems.

The underlying question is straightforward: can we design systems in which responsible behaviour is supported by architecture and process, rather than relying exclusively on the intentions of the people who build them?

This is a question of implementation as much as philosophy.

[Link to: Ethical.ink — AI safety, risk and governance research]

[Link to: Samurai Labs — Research and development]

Bigger is not always better

The pursuit of greater computational capability can obscure another question: how efficiently are we using the capabilities and resources we already have?

Consider an AI system asked to retrieve an obscure fact that is absent from its training data. A larger model may still need to search for that information. If two models ultimately retrieve the same answer, the additional computational expense of the larger system needs to be justified by a measurable benefit.

This is the intuition behind my Zhu Zhu bird thought experiment: a larger model can consume substantially more resources without producing a better result for a particular task.

The point is not that larger models are inherently inefficient. It is that model size, task performance and economic value are different variables.

We should evaluate intelligent systems according to the outcomes they deliver, the resources they consume and the wider value they create.

[Link to: Factory-Floor Thinking — resource efficiency and systems optimisation]

[Link to: Knowledge on Demand — shared information infrastructure]

Beyond the individual model

The next stage of AI development will involve increasingly interconnected systems: agents that retrieve information, use tools, exchange data and coordinate with other agents.

This creates opportunities for greater capability, but it also makes system-level design more important.

How should information be shared? How can provenance be maintained? How should permissions and responsibilities be distributed? How do we ensure that individual components can be monitored without losing sight of the behaviour of the whole?

These questions connect AI safety to information architecture, semantic interoperability, cybersecurity and organisational governance.

My wider research explores how shared knowledge infrastructure and clearer semantic relationships might reduce duplicated work while making information easier to retrieve, compare and evaluate.

The objective is not simply to build more intelligent systems. It is to make intelligence more useful, more accountable and more efficient.

Technology, humanity and the future

AI safety and responsible technology are not separate from the wider questions facing humanity.

The way we develop intelligent systems affects employment, access to knowledge, the distribution of resources, environmental pressures and the relationship between human judgement and automated decisions.

We need to consider these effects together rather than treating every technical achievement as an isolated success.

I believe technological progress and responsible control can reinforce one another. Good safeguards can support trust. Better information architecture can reduce waste. Clearer accountability can make innovation more sustainable.

The challenge is to build systems that serve a purpose beyond their own expansion.

We do not have to choose between technological progress and responsible control. We need to design for both.

Explore further

• Ethical.ink: Research and practical thinking on AI safety, ethics and governance.
• Samurai Labs: Research, development and systems design.
• Shared Values for a Better Future: The philosophical principles informing my approach to technology and humanity.
• My Journal: Explorations of ideas, questions and connections across disciplines.