“Data shortages may not matter so much if computer scientists can develop new techniques or architectures that make models learn more efficiently on less information.”
– Katyanna Quach
Current AI models have an apparently insatiable appetite for data, usually in the form of text. They draw their inferences based on the data in their training sets, and typically, the bigger the training set, the more accurate the model.
However, the AI training demand is growing faster than the available text supply.
Researchers are looking for additional and alternate sources of text, facing forecasts of a few months to a few years before the supply of text is exhausted, at least in its traditional form.
Is there enough text to feed the AI beast? | SEMAFOR | April 16, 2024 | by Katyanna Quach