On Feb. 25, I learned how Barbie gets her voice during the final installation of MakerWeb’s Hello Barbie Teardown campaign.
Mattell’s “Hello Barbie” toy, which MakerWeb has waged a small but influential war against, is a speech recognition device that has the ability to pick up key words in dialogue and maintain conversation with the child speaking into it.
Makerspace Coordinator Amanda Ervin dedicated the final lecture of the teardown campaign to helping students understand how this kind of technology is possible, and furthermore how it could be potentially altered.
Although Erwin accepts that, like any technological device, Hello Barbie will never understand empathy no matter what speech material you program it with, she believes that changing what Hello Barbie says can still move mountains in the world of toys.
“The ultimate goal? New Hello Barbie,” Ervin states.
Although we will never be able to program Barbie to understand emotion, altering her conversation material enables people to change her ideals, and even more importantly, the stereotypes she embodies.
“I want people to say: ‘I know that I can change it. I see myself as part of the solution,’ ” says Ervin.
So how can this actually be done? Here is the basic idea:
It begins with what is called a logic chip, where all of the data and information is stored. When a microphone is attached to the logic chip, speaking into the microphone transmits audio into voltage.
This voltage is produced in the form of signal waves, which the logic chip transmits into actual sound.
Text can thus be output through serial communication, which is put into a larger computing system, predominantly a small computer called a Raspberry Pi.
This Raspberry Pi, or a tiny version of it, is installed inside each Hello Barbie.
Taking a step deeper into the mechanics behind Hello Barbie, Ervin proceeded to explain how Barbie got her memory and how she knew how to respond to different dialogue.
As an explanation, Ervin showed us playground.pandorabots.com, where anyone can make his or her own “Chatbot.”
These “Chatbots” can be trained to answer specific questions with specific answers. For example, if you type in “Hi” to your chatbot, you can change the bot’s answer to “Hi, how are you?” You can even train your bot to have multiple answers to questions, and to pick up on key wording so that it responds appropriately during conversation.
As a group, we played with the chatbots for most of the seminar, amused at the intricate training of the chatbots.
“I would love it if people got really creative about it,” Ervin says. “This is about what they would want that child to have a conversation about.”
Ervin is posting tutorials on all segments of the campaign, from how to mold your own Barbie to reinstalling new chatboxes inside of them.
The tutorials can be found at muse.union.edu/makerweb.
Amanda Ervin will provide any and all materials necessary to participate. To get materials, email her at firstname.lastname@example.org.