Proliferation of Franken-Algorithms

Global Rush to Fund A.I.
March 10, 2020

Proliferation of Franken-Algorithms

While a single algorithm might be easily described and deployed as expected, systems of algorithms all working together can sometimes pose problems.

Algorithms are simply rules that define and automate the treatment of data. They are built using “if this, then that” logic that a computer can understand and process.

Here’s an easy example: If a website reader’s IP address is based in Baltimore, the rules allow that reader to freely access the site; if the IP address is based in Belgium, the rules first show a GDPR screen stating privacy and cookie policies.

While a single algorithm might be easily described and deployed as expected, systems of algorithms all working together can sometimes pose problems.

Developers don’t always know in advance how one algorithm will function alongside other algorithms. Sometimes, several teams of developers are all working independently on different algorithms and data sets and only see their work once it is deployed.

This has been the cause of recent stock market glitches and e-commerce website wonkiness.

Indeed, it is a challenge for big companies like Facebook, which have billions of algorithms working together at any given time.