Google uses its Global Impact Awards to award grants to small institutions for whom a mere $1.2 million (“mere” when you compare it to the full extent of Google’s wallet) is a huge impact, and places that are also unafraid to “fail fast or challenge the status quo, and have what [they'd] like to call ‘a healthy disregard for the impossible.” So it both makes sense and is pretty cool that Google would choose to give $1.2 million to the Geena Davis Institute on Gender in Media and access to their engineers in order to craft experimental software that can analyze gender representation from raw television footage.
The GDIGM’s last study, reasearched manually on 129 family films, 275 primetime television shows, and 36 kids shows featuring nearly 12k speaking characters, yielded pretty significant results. Like that only 28% of speaking roles in the films had gone to female characters, or that representation of women in science in fictional media is even worse than it is in real life. If the Institute was able to create software that could use voice recognition to accurately analyze speaking roles, it would free their research partners to do more analysis instead of data collection. It might even be possible to have analytics available the day after an evening of primetime shows air, much like the turnaround of Nielson ratings.
Whether the technology can effectively achieve those goals is yet to be determined: creating the software will be at least as experimental as it is practical. But, as Google’s guidelines for its Global Impact Awards state (or at least they would if they were said through one female character in a kids’ show that comes to mind) taking chances, making mistakes, and getting messy is the name of the game.