# What Is an AI Anyway?

A personal reflection on digital species, awareness, and the future of intelligence.

For the longest time, I thought of AI as math, algorithms, and software. That was how I looked at it, as engineering, logic and machine learning. But recently, given all that is happening with AI in the world and a TED Talk I watched altered that framing. The speaker posed the question what is an AI, really? and then described it as something I hadn't heard before:

A digital species.

It landed. Suddenly, I wasn't just thinking about code and math anymore. I was thinking about life. A kind of life that lives on servers and infrastructure, learns from data, and interacts with us humans, across screens and systems. Something we shape, but that in turn shapes us.

Having previously worked in defence command and control, I have seen a few different types of AI applications, agents, and predictive models deployed, tested, researched, and visualised. I have always had a sense of practicality with AI. I would ask, how can we visualise this so the user doesn't need to understand the complexity, yet can still harness its power instantly and intuitively?

Now, as someone building technology every day in a startup, this digital species description made me reflect on the kind of AI we're actually trying to build. Not just technically, but philosophically.

AI: Math But So Much More

To be clear. AI is math. Neural networks, weights, data. It's software too. But that doesn't tell us what AI is, only what it's made of.

Recently, I've gone down the AI rabbit hole. Watching TED Talks, listening to hours of Lex Fridman's podcasts, and watching materials from Sequoia Capital's AI Ascend 2025 series and more. Across these, there's a recurring thread. No one fully agrees on what AI is, but everyone agrees it's more than just code. It's becoming something we live with, learn from, and increasingly collaborate with.

Today's AI, especially generative AI, feels like more than math. It can write, illustrate, animate, converse, and even joke. It can take inputs like text or images and create something new. One of my current new favourite things is watching the generative AIs print out math.

NVIDIA puts it well:

"Generative AI enables users to quickly generate new content based on a variety of inputs… [using] neural networks to identify patterns and structures within existing data to generate new and original content."

Generating new content is an awesome way to describe the transformation that AI can do to data. But it's not the only transformation. Sometimes data gives you value not through answers, but through awareness. Not through prediction, but through perception.

Generative AI models are trained using supervised, unsupervised, or semi-supervised learning techniques and increasingly on unlabeled data. That's a big shift. These models learn on their own, then produce content we can understand, act on, or be amazed by.

It's not a tool like Excel. It's more like a system that learns and evolves.

AI in Business

One of the most exciting frontiers of AI is its application in business. Echoing Sequoia Capital's AI Ascent 2025, we believe the real opportunity lies in the application layer. While massive innovation continues in the backend, the models, infrastructure, and compute, it is at the application level where business value is unlocked.

From startup velocity to enterprise scale, we now have the ability to speed up time to value. AI lets us iterate faster and create outputs that are not just functional, but intuitive, elegant, and even joyful to use. When paired with great data and thoughtful design, these AI outputs generate real-world impact, from cost savings to risk reduction to new revenue streams. We're no longer imagining if this is possible. We're building it now.

Enter AI Agents

What we're seeing now is a step beyond traditional language or image models. Multimodal AI is the new frontier. These are systems that can interpret text, images, audio, video, and even sensor data at once. These models do not just generate language or visuals in isolation. They understand context across media types. That's closer to how humans think. We hear, see, read, and feel at the same time to understand a moment.

Imagine an AI that can watch a surveillance feed, read a maintenance report, and process a voice message. All to help determine the operational status of a factory. This kind of multimodal reasoning is already emerging in research and starting to appear in real products.

In parallel we're also witnessing the rise of Agentic AI. These are systems that not only understand but also act. These are not just assistants. They are agents. They learn from interactions, pursue goals autonomously, and make real decisions. Not in the science fiction sense (but how awesome and scary would that be), but in very practical ways. Like dispatching an alert, generating a report, or flagging a risk before you ask.

This is no longer just AI that responds to us. It is AI that collaborates with us.

Data to Awareness

The common view is that data becomes insight, which drives decisions, which lead to actions. It's a neat little value chain:

Data → Insight → Decision → Action

But that's always felt too linear and also like the default status quo way that everyone in business looks at it. It is valid and technically how things could happen. Real-world decisions are not always based on data. Sometimes the value lies not in deciding, but in being aware. In seeing what's happening. In knowing what might happen. Or in choosing not to act, which is still a decision. So we see that business value can be derived at all stages of value chains.

So I've come to prefer a different framing: Data → Awareness

Awareness is a richer space. It allows perception, comprehension, and projection. It allows you to say, "I understand this situation, and now I know whether or not I need to respond." That's more aligned with how people and organisations work. Across finance, impact, operations, sustainability, security, and beyond.

My contrarian view is that data is only a percentage of the truth and that, in most cases, it does not need to be 100% to get to a level of awareness that creates understanding and support and thus value.

Awareness: Why We Building

I co-founded a company called The Awareness Company. The name wasn't just branding. It was the philosophy. We saw the power of situational awareness, especially in fields like mining, infrastructure, and conservation, where real-time visibility changes lives and saves them.

Our product, HYDRA, started as a way to digitise and productionise that kind of awareness, the kind that we saw during our time in defence. It automates the creation and integration of data, aggregates it across sources, and generates insights and intelligence in real time. It is built on an ontology, a structure that defines how things relate, to turn fragmented business inputs into meaningful signals.

We built HYDRA because we believed that good data enables useful AI. Without structured, real-time, clean data, even the best models will produce noise. And if AI generates outputs from noisy data, the harm is worse than having no AI at all. Time is wasted. Trust is lost. Decisions go wrong. That's why I always say: Good data = happy AI.

Neural Networks, Learning, and Cognition

Before diving into how AI works with data, we need clarity on two foundational ideas: ontology and machine learning.

Ontology, at its deepest level, comes from philosophy. It is the study of being, a branch of metaphysics that seeks to understand existence itself. In the context of AI and data systems, an ontology is a structured set of concepts and categories within a domain, along with their properties and relationships. It gives AI a map of meaning, a way to understand how "driver," "vehicle," "incident," or "location" relate in a system like HYDRA.

Machine learning, as Coursera puts it, refers to the general use of algorithms and data to create autonomous or semi-autonomous systems. Within that, deep learning uses layers of neural networks, structures inspired by the human brain, to perform increasingly complex tasks. These algorithms don't need explicit rules. They learn patterns from data.

This brings us to aggregation and insights. In the real world, data is highly fragmented. The first challenge is to create and integrate that data, from apps, sensors, or legacy systems. But the second challenge is to structure and aggregate it automatically. That is where ontology plays a central role.

In HYDRA, our brain is ontology-driven. It helps unify data from different sources, turning raw fragments into internally labeled datasets. These can be used by AI models to generate real-time insights, track change, and reveal patterns across systems.

Foundation Models, Everywhere

The evolution of foundation models like GPT-4, Gemini, Claude, and open-source equivalents has accelerated.

They are used not only for text generation, but for:

These foundation models are not just passive engines for generation. They are becoming more context-aware and capable of autonomous behaviour, a shift that leads directly into the rise of agentic AI.

We are living in an age where these models are increasingly customizable, context-aware, and often task-specific. This opens new ground for AI in business environments like logistics, utilities, compliance, or incident response. And more so, they are changing at a never before seen pace.

Agents in Organisations

Most business intelligence tools today help users see what happened. The more advanced ones help predict what might happen. But what we're moving toward is something deeper. Systems that know what's happening right now and can act on it.

This is where platforms like HYDRA or other operational intelligence layers can evolve. Not just surfacing real-time information, but integrating multimodal input like video feeds, text logs, and sensor data, applying ontology-driven reasoning, and triggering autonomous responses.

For example:

The future is not just about alerts. It is about systems that interpret, prioritise, and act in real time.

AI as a Digital Species

So let's return to the idea that started all this. AI as a digital species.

It learns. It creates. It evolves. It lives in our systems and workflows. It does not sleep. It grows stronger with data. It becomes part of our digital ecosystems, sometimes as a co-pilot, sometimes as a shadow.

Framing it this way makes us more intentional. It forces us to ask: What kind of species are we designing? What values are we encoding? What boundaries are we drawing? Is there a line?

At The Awareness Company, we don't want AI to replace people. We want it to make people more aware, more intentional about being better with data and AI. More capable. More informed. We want to help people, from the boardroom to the field, become heroes for the planet, for people, and for profit.

Not because AI tells them what to do, but because it helps them see more clearly, across their operations, their impact, and their future.

To the future.