No knowledge, no AI? If you want meaningful results from an AI or ML project, then make sure your data house is in order – data, metadata, content & their curation – otherwise you are just fuelling your systems with your past mistakes. Of course, it may be an expensive way of doing it, but launching an AI project will also draw attention to data quality or the lack of it, when questions don’t produced meaningful answers.
What is the point of hiring lots of data scientists when you don’t ‘know’ your content? What is the point of hiring lots of (scarce) data scientists when they are going to spend 80% of their time sorting out the data before they can actually do any ‘science’? If you had ‘the knowledge’ these problems would be substantially mitigated:
data is badly labelled so the machine ‘learns’ from the wrong information;
ML & AI need data that is unbiased & of sufficient breadth, equally distributed across target input ranges to avoid further bias;
what metadata does the business collect and why?
ML & AI require sufficient data, a meaningful sample, in order to generate legitimate results;
those who create data need to ‘know’ what data the data scientists need for AI or ML;
poorly calibrated instrumentation results in poor data, faulty measurements or needs to be made canonical;
process rigidity or complexity generates data and application model opacity;
data uninformed by semantics, ignores the 80& of information held in unstructured content;
data quality needs to be an ongoing task to avoid putting new ‘bad’ data into ML models.
Knowledge graphing is a necessary element of getting the most out of your AI or ML:
with knowledge you can decide whether you have the right data & right amount of data to be able to task your data scientists;
machine learning acts as a magnifying lens – small inconsistencies become much larger ones – keep on curating;
machine learning is iterative, so keep an audit trail between data iterations so you learn from your mistakes;
make sure your content is owned & curated by subject experts from the business, distinct from the AI or ML team;
involve the business experts to capture their expertise and build that into your knowledge.
If you gain knowledge about your organisation, you will have better-informed strategy; decision-making; objectives & tactics.
Imagine if Amazon had been unable to tell the difference between ‘apple’ and ‘Apple’…..
