Data Testing and AI - "What if"

A bit of a preface before diving in. I am not a data scientist nor do I have the patience to be a data scientist. So what are data scientists. First from my experience they analyze data for companies and make the final determination on everything from is the data "good", "accurate" and /or "usable" or not. Much like a home inspector or credit rating, etc. they are paid to be skeptics about the data and to find problems with it. On the easy side missing fields in data, inaccurate details, or outdated details and more are often easy to find and may completely discount a whole database from being fit to purchase. But past the obvious is where the water becomes murky.

It seems there are as many different testing methods as there are data scientists and each are 100% completely accurate, guaranteeing their data is the best. There does not seem to be an industry standard though many use companies like Truthset to measure their data against other data sets to judge their overall data quality.

Based on all of this are we really ready to say any data is ready for AI? We know early on AI has been fraught with inaccuracies and caught putting out information that has been troublesome for some people based on false information about them that somehow made its way into the AI programs path.

Now take another leap to marketing using AI to help build audiences, target customers, create campaigns based on audience information and data and it does not seem far fetched to believe some of these campaigns and delivery could go seriously sideways, not only targeting the wrong audiences, but costing companies millions in lost marketing budget.

Anyone who buys data or uses it for sales and marketing related activities are typically eager to share stories of data gone wrong. While its not like making a mistake during brain surgery, the repercussions of a data mistake or misuse could have long term repercussions for the consumers it affects.

What is the solution? First we need to have or set some kind of standard testing method that we deem as accurate in order to compare apples to apples when evaluating data. Then based on these standards how well do those standards and methods guarantee the data accuracy? From that we can then determine if it is accurate enough and ready to be used in AI. While AI is supposed to correct for human error it is still only as good as the information it has at hand, the accuracy of which is determined by humans.

Another way of looking at it is do you trust engineered food? Would you ingest it even after testing and reports based on the testing? The same can be said of most if not all data which is engineered. Until there are quality standards and it is deemed fit to ingest then maybe taking a step back from AI and using it in marketing and a number of other areas is warranted

Thoughts?.