Machine Learning Services: To Develop Inhouse or To Outsource / Lessons from History

Until 10,000 BC farmers cultivated plots just large enough for sustenance for their families. Subsequently with the emergence of great civilizations, leaders thereof encouraged organized agriculture that could feed their whole population.

A few centuries ago individual communities, dug wells, harvested rain water, erected hand pumps to draw water for hydration and sanitation. Today all of our water is pumped and distributed by utilities.

Similarly, many people in the late 19th century were leveraging the principles of electromagnetism to generate their own electricity but circa 2017 Con Edison is most efficient at generating it for all of us.

 Incidentally, numerous companies incorporated to provide internet services when the telecommunications technology was in its nascence but today we only have a handful: Verizon, Comcast etc.

Finally in the late 90’s firms across industries were building their own datacenters to store the reams of data their operations were generating. Today everyone from clandestine organizations to medical imaging startups to genomics firms to Drug Discovery enterprises store their data on the cloud whose infrastructure is maintained by a few conglomerates like Amazon, MS, Google, IBM etc.

Economies of scale are the invisible forces guiding the markets for each of these technologies from an initially fragmented state towards an inevitably consolidated one, as must come as anything but a revelation. Notwithstanding it forced me to revisit a post I recently authored where I drew parallels between Electricity and Machine Learning and proposed that healthcare firms should work towards incorporating the latter into their Shared Services.

My logic was its better to develop ML prediction capabilities inhouse as

1)     The needs of organizations are unique and ML solutions today are not adaptable

2)     It allows organization to hone their ability deploy cutting edge technology

But after reflecting on the path charted by past disruptive technologies it seems historically a handful of firms have been able to take solutions like ML to scale and meet the needs of participants of a diverse economy, no matter how unique. Perhaps then it is more prudent for firms train their employees or hire those with the skills to use these external Machine Learning Services/Tools rather than build their own.

What do you reckon? Love to hear your thoughts