T.E.P. 's comments about " supplying picks and shovels" to big...

T.E.P. 's comments about " supplying picks and shovels" to big AI users reminded me of my own long-standing view of APX's position in the eco system of AI/Machine learning industry.

The analogy I'd use is China in the global supply chain as the manufacturing factory for the world. Yes it is very labor-intensive. Yes the margins are "miserable" compared to its many customers with more bargaining power (e.g. Apple). Yes, it doesn't look very tech savvy. However it has generated tremendous wealth over the last few decades just by doing those tedious and low value-added manufacturing works thanks to the sheer volume of orders. And it has made himself an indispensable and hard-to-replace partner for many multinational brands, due to its huge size of labor pool, excellent infrastructure facilities and pro-business political and social systems. These are the most prominent comparative advantages / economic moats of China. Developed countries have long been talking about reducing supply side risks by either bringing back manufacturing to their domestic domicile or to shifting it away to other developing countries. However, the results so far have not been that encouraging.

In many ways, APX shares similar traits to China, though in different sectors. The crowd that APX has been building through out the years is certainly the most valuable asset that it has. And they've been constantly improving their technology and crowd management systems to bring out more productivity: higher quality of data, quicker responses to clients requests, etc. This has made APX the favorable partner for the biggest tech companies in the world for many years and will likely remain that way for many years to come, as long as their customers need high quality and the most updated data to sharpen their own edge vis-a-vis the mounting competition.

When it comes to the question of why tech giants would not build their own data labeling service, I think there are two main reasons: 1) the low margin, associated with the huge workload of managing millions of crowd workers, makes it unattractive ; 2) data security/exclusivity will be a very sensitive issue for tech companies to consider when sourcing data from a competitor.

ALL IMHO.