I’m seeing a lot of "AutoML" tools that claim to do everything from cleaning data to picking models. For those in the field, do you think we are at risk, or is there a layer of business context and ethics that these platforms simply can't touch yet?
3 answers
Tools like AutoML are great for the "how" of data processing, but they are useless at the "why." A Data Scientist provides meaning to the numbers. They understand the nuances of the business problem, the bias inherent in the data collection process, and the ethical implications of the results. Without a human to interpret and question the output, you risk making very expensive mistakes based on technically "accurate" but contextually irrelevant data models.
If the model is more accurate than a human-built one, does the "why" actually matter to a CEO?
Automation will take the "science" out of the entry-level tasks, leaving the "strategy" for the pros.
I agree with Michelle; the "drudgery" of data cleaning is what goes away, but the high-level decision making is staying firmly in our hands.
It matters immensely, Steven, because a model might be "accurate" by identifying a correlation that is actually a legal or ethical liability. A CEO needs to know if a model is discriminating against a protected group, something an automated platform might not flag unless a human specifically tells it to.