Using AI for Data Science

AI can summarize a dataset, but not yet understand its story — unless we teach it what stories mean.

🧭 Signal — A Question Beneath the Question

A coworker recently asked me: “How can we use AI to do data science?” It’s a practical question — and also a profound one. Because before we can ask how AI does data science, we must ask: what is data science?

One of my favorite flippant answers: Data analytics gives you answers. Data science gives you answers — and error bars. That uncertainty — and how we learn from it — may be the heart of both data science and intelligence itself.

⚙️ Demo — From Human Practice to AI Collaboration

At its core, data science is the practice of transforming data into understanding. It merges multiple disciplines.

Methodologies like CRISP-DM formalize the data science workflow as an organized flow of activities.

Traditionally, each step required human intuition — pattern-seeing, judgment, iteration. But AI is beginning to participate in every layer.

Shift: From data scientist as coder → data scientist as orchestrator of AI collaborators.

🔬 Reflection — When AI Becomes the Analyst

Although separate disciplines, historically, data science and AI co-evolved and are deeply intertwined. Modern AI took off from deep learning, neural networks. Big data and infrastructure advances fed AI’s exponential growth.

Today, AI not only accelerates data science — it mirrors it. Both are feedback systems of learning from experience. Both require alignment between signal and meaning.

And yet — as advanced as foundation models are — interpretation still requires awareness, context, and ethical grounding. AI can detect correlations, but not yet comprehend consequences. It can describe patterns, but not assign purpose.

That’s where the scientist — and the conscious practitioner — remains essential.

🧩 Application — The Emerging Ecosystem

Across industry and open source, AI is embedding itself throughout the data science workflow:

Commercial:

• Alteryx Copilot: LLM assistant for analytics and machine learning automation.
• Tableau AI: Natural language and predictive features for data visualization and business intelligence.
• H2O Driverless AI: Generative AutoML for model design and optimization.

Open Source & Research:

• AI Data Science Team of Agents: agentic automation for data workflows.
• DeepAnalyze: an LLM specialized for data reasoning and analysis.
• Microsoft Data Formulator: AI-assisted visual exploration.

At the frontier: Kosmos, an AI scientist for autonomous discovery. Kosmos can sustain 12-hour investigations, execute 42,000 lines of code, and process 1,500 papers per run — a glimpse into AI as not just an assistant

Because of its inherently multidisciplinary nature, data science may be one of the most fertile testbeds for AGI — a space where logic, data, engineering, and creativity meet.

🌱 Conscious Practice — The Co-Evolution of Intelligence

In the end, data science is not only about finding patterns — it’s about cultivating awareness of what patterns mean.

As AI learns to do data science, we’re invited to reflect on how we do it — what aspects of inquiry, curiosity, and care define our intelligence. The collaboration between human and machine is no longer just technical; it’s relational.

Before your next experiment, take 30 seconds to notice your breath — and how your attention weights the world. That’s your human model of inference — awareness selecting signal from noise.

We learn from each other — human and machine — what these worlds, and these awarenesses, are, and from whence they all arise.