5 Brutal Truths Data Scientists Won’t Admit About AI

Read time: 2.5 minutes

5 Harsh Realities for Data Scientists to Grasp

1. AI IS CAPABLE OF WRITING YOUR CODE

Python.
SQL.
Models.
Pipelines.

Coding is becoming increasingly automated. Your MOAT is not your coding skills; it is your ability to make decisions.

Knowing why and how to create something and knowing when not to create it is much more important.

2. YOUR MODEL IS NOT YOUR MOAT

Models are getting more and more accessible.

Your competitive advantages are: better questions, better data, better context, and better decisions.

3. AI MAKES YOUR BAD DATA EVEN WORSE

Bad data = faster analysis.

Bad assumptions = faster predictions.

Bad labels = faster mistakes.

AI does not recognize the errors in a given source of information, and it can multiply your errors.

4. ACCURACY DOES NOT MEAN IMPACT

The model can be accurate at 99%, and still bring no value for the business.

If no decision is actually changed based on the model, one should think about it, and ask themselves “What’s the point?”.

The measure of success is not just how good the model is; it is the effect it has.

5. DATA SCIENCE IS TURNING INTO SYSTEMS SCIENCE

Old thinking: DATA → MODEL
New thinking: DATA → CONTEXT → AI → DECISION → ACTION

A model is only one piece of the system. The valuable data scientist understands the entire chain.

💡Key Takeaway: 

AI is not putting an end to data science. Rather, it is taking away some of the technical barriers.

In fact, successful data scientists will not be those who can code the best.

The successful data scientist will: Define the correct problem; Assess the data; Question the assumptions; Impact decision making.

AI can design the model. But you still have to determine what changes the model must make.