algorithm; ended up removing the tag “gorillas” ▸ “Machines aren't biased, but they can easily learn racism from us if we aren't careful.” - Yonatan Zunger, Chief Architect of Ambient Computing at Google ▸ Make small improvements to algorithm around language and skin tone ▸ Expand data set & get free retraining labor from the affected
ECONOMY? ▸ More granular software that can be easily implemented into other projects; similar to micro service and SOA architecture ▸ More cutting edge algorithms on the market; fewer stuck in academia and lower barrier of entry ▸ Standardization through reuse and chaining ▸ Better quality assessment ▸ External validation & assessment
have access to many of the same “big data” tools ▸ Hadoop, TensorFlow, Spark, ElasticSearch ▸ “Big data” companies are not big data by design, they became big data in the course of pursuing insights into their service and customers ▸ Similarly, those on the edge of algorithmic design are doing so to further another product
WHAT ABOUT BIG DATA? IF BIG DATA TOOLS ARE OPEN, WHY NOT ALGORITHMS? ACCESS TO COMMERCIALIZED ALGORITHMS INCENTIVIZES ACCURACY & VALIDATION CHOICE IN ALGORITHMS CAN REFLECT BIAS OR VALUES