Beyond ChatGPT:
How Data-Driven AI Is Reshaping Repetitive Work
Data-driven AI goes far beyond language models like ChatGPT. Learn how AI is evolving into a tool that can model and replicate complex human tasks, and why this matters for industries like construction, food service, and care.
AI Is More Than Just ChatGPT
Most mainstream discussions about artificial intelligence centre on Large Language Models (LLMs) such as ChatGPT. But these conversational systems are only one part of a broader field.
At its core, data-driven AI is a collection of numerical methods that model real-world processes, physical or abstract, by analysing vast amounts of data. This approach is rapidly changing how we think about work, automation, and the replication of human skill.
What Is Data-Driven AI? A Mathematical View
Data-driven AI doesn’t derive rules from first principles or human logic. Instead, it creates nonlinear partial differential equations by learning from thousands, even millions, of input/output examples.
Causal Modelling with AI
Processes in AI are considered causal systems, given the same input, they should always produce the same output.
AI “learns” the structure of these systems from observational data, not underlying logic.
This is a heuristic approach, where the goal is not to find the “true” rules, but to approximate behaviour well enough for reliable prediction or replication.
From Language to Labour: How LLMs Led the Way
LLMs were originally designed to tackle translation tasks between languages. But it was soon realised that:
Questions and answers can be viewed as two different languages.
This insight opened the door to applying LLMs to general reasoning tasks.
This breakthrough popularised AI as a tool for solving abstract problems. But the next frontier is more physical.
Process Modelling: The Next Wave of AI
Stochastic Calculus and Uncertainty
The training of data-driven AI resembles stochastic calculus, the mathematics of randomness. This is why AI struggles with:
Edge cases (rare combinations of inputs)
Non-causal outcomes (where inputs don’t fully determine the result)
These weaknesses must be addressed before we approach Artificial General Intelligence (AGI), but even current AI models are robust enough to describe many repetitive human tasks.
Real-World Application: NVidia Isaac GROOT
NVidia’s Isaac GROOT is a new software platform that uses data-driven AI to teach robots to emulate human tasks in physical environments. It applies AI modelling within the constraints of our three-dimensional world.
This allows for the emulation of:
Dexterous, repetitive tasks
Skilled labour workflows
Voice-command-driven actions
AI and the Automation of Manual Labour
Why the Commercial Motivation Is Strong
Just three sectors, construction, food preparation, and domiciliary care, employ over 25% of the UK workforce.
Many roles in these sectors involve:
Repetitive manual actions
Simple procedural logic
Skills that can be taught by example
AI-powered humanoid robots could transform this dynamic.
A Future Scenario: AI in the Kitchen
Imagine a restaurant chain where:
A head chef designs new recipes.
These are demonstrated to a robot using GROOT-style learning.
Within a week, the recipe is rolled out, robotically executed across all locations.
Benefits include:
Zero staff training required
Instant, exact replication
Reduced cost and human error
Apprenticeship by Observation: How AI Learns Human Skills
This vision mirrors traditional apprenticeship learning:
Observe an expert
Repeat the task
Refine the skill over time
AI can now replicate this structure by training on sensor-rich data from human demonstrations.
Key Limitation: Dexterity and Non-Standard Outcomes
Despite progress, the main challenges remain:
Replicating human dexterity
Handling unpredictable situations
Until these are solved, AI will be domain-limited, excelling in repetitive tasks, but struggling with nuanced, creative, or judgment-based work.
AI as Process, Not Just Language
AI is evolving from answering questions to performing actions. Data-driven AI is not only modelling thought, it’s modelling labour.
In the near future:
Robots will learn by watching humans.
Physical tasks will be automated with AI-built models.
Sectors with repetitive jobs are poised for radical transformation.
The UK must start thinking about AI process modelling as a workforce issue, not just a tech trend.
