Neural Toons Original

Your AI Can't Tell Cause From Effect

28 May 2026 · The Neural Feed

“Causal discovery: still cheaper than a PhD, but not as cheap as a spreadsheet.”

The story behind the toon

What happened: Causal discovery: still cheaper than a PhD, but not as cheap as a spreadsheet.

This Neural Toon is based on the story “Why LLMs Fail at Causal Discovery and How Interventional Agents Escape”.

Panel 1 – Zoe: Leo, my chatbot just recommended umbrellas because it saw rain in the forecast. That's not causation!

Panel 1 – Leo: Causation? I see a 40% cost saving if we switch vendors.

Panel 2 – Zoe: LLMs can't do causal reasoning — even after $50,000 of fine-tuning. They only see correlations.

Panel 2 – Leo: So the $50k we spent last quarter fixed nothing?

Panel 3 – Zoe: Correct. But here's the kicker: A-CBO can match fine-tuned models with zero training cost.

Panel 3 – Leo: Zero training cost... I can fire our 3‑person data science team.

Panel 4 – Zoe: Except A-CBO only works if you tell it which variables to test. Otherwise it's guessing blind.

Panel 4 – Leo: So we still need someone who understands the business. Great.