Why is AI still in the data-rich industry domain .... ?

How long it will take for effective AI deployment across range of industries both massive data-rich and data-scarce ?

We see that AI in the present day revolves around data companies like Facebook, Google, Microsoft, Whatsapp, Amazon etc.. Let us try to understand why is it that only these industries are able to grow and apply AI where a whole lot of other industries from manufacturing, helthcare, transportation, defence, government etc are limping for want of quality AI apps. How can we find the right environment for AI apps to thrive and grow.

The answer lies in the software centric AI sysytems that depend on the quantity of data to process. We brought terms like Big Data to annopunce to the world that whatever be the data, the quality compromising, we need lage amonts of data running into Giga bytes to be processed for the success of AI applications. And the sad fact was that we got this data only with data rich applications that generated lots of data, like data processing, email processing, information search, retail sales, etc. 

But it is a real fact that there are other sectors like manufacturing, healthcare etc which also need to reap the benefits of AI but do not have the capability to produce enormous amounts of data like data processing, emailing, search, retailing etc. 

The focus now needs to shift from software centric approach in AI development to  data centric approach. AI apps should be in a position to be used to reap benefits even with small amount  of high quality data or data that are very accurate and precise which otherwise we get from from experts in the concerned manufacturing or other data scarce industry areas. 

Research should continue in areas where we are able to compromise for the quantity of learning data, by having an improved measure of the quality of the data which we use in this learning data set. Very less high quality data should be able to suffice or replace for large quantity of questionable low quality data. Then only can we think of getting AI implemented and popular over all sectors of the economy. Banks can think of AI adoption as they are able to generate large amounts of data (big data) unlike an educational institute or manufacturing cos.

Future AI research should focus in this data scarce, high quality area then only can we reap the benefits of AI across sectors. Research by Andrew Ng that appeared in HBR of July 2021 titled AI doesn't have to be too expensive or complicated for your business (click here) was helpful in this writing. 

Moving from a software-centric approach to a data-centric approach in developing future AI systems appears fine, how do we achieve it ?

George