Scale AI to Build The First All-Baby Data Annotation Workforce

This week Scale AI and Meta Inc revealed a deal that will heat up the AI for Babies race immensely. “ChatGPT for Babies,” is just over the horizon and the public’s anticipation is at an all-time high for the product that will “allow every baby to flourish.”

A BTO insider from Scale AI told us that to make the AI baby product safe and accurate, some 100 million ‘oos’ and ‘ahhs’ will need to be labeled by real babies, preferably somewhere cheap like the Philippines. This includes both pre-training data-labeling, and evaluation of AI generated coos executed by thousands of freelance babies.

“It’s really important that our baby taskers are highly educated in their babbling, crying, and laughing,” said Alexander Wang, Meta’s Chief of AI. “Scale AI is a premier partner in picking subcontractors who pick good subcontractors who pick good babies.”

However, critics say the data labeling work is uniquely taxing, and AI product comes with significant costs. BTO spoke with one mother about how her infant child must stay glued to the screen to support her doll house of 5. “Every minute of the day, she shakes green or red rattles, and at night she wakes up with the coos of ghostly babies in her head.” The pay, however, is the best she can get despite her impressive resume – fluent in eye contact and wave hi.

Scale AI’s CEO says their baby data annotation workers live exciting lives. “We work dutifully to ensure the mental health of our babies are cared for. As soon as a baby shows signs of depression or suicide we intervene with intensive therapy and support.”