The ultimate goal of Generative AI is to automate repetitive human routines and emulate humans' creativity by producing original artworks. In short, to automate humans' jobs.
But before we are able get there, researchers, data scientiests, data engineers, machine learning engineers, software engineers and even testers need to put in tremendous amount of manual labour to source the right data, clean up the data, correct the data, label the data, and train the AI models by using carefully curated data, test the AI models, retrain and retest the AI models using newly discovered production data.
In fact, we could say that Generative AI is the outcome of the mapping of human intelligence to machine intelligence via repetitive manual labour.
And the irony is during this engineering journey, Data Scientists, Engineers, and Data Labelers are steadily using manual labour to automate themselves out of their jobs.
----
This newsletter by VentureBeat (The human side of generative AI’s gold rush) has revealed another paradox that happens behind the scene for the current hype surrounding Generative AI - low-paid data labeling labourers are employed to label the training data for Large Language Model (e.g., ChatGPT) in order to make the AI model more intelligent.
Posts
0 Comments