You can learn them all. But will it?
Large language models are amazingly good at understanding and generating language. However, there is an often overlooked bias against languages that are already well known on the Internet. This means that some languages may succumb to major technological advances in AI.
Researchers are looking at how it works and how it shifts the balance from these “resource-heavy” languages to languages that haven’t yet made a big footprint online. We spoke to several researchers who are trying to make languages like Catalan and Jamaican Patois more accessible for his AI language model. Their approaches range from creating original datasets, to examining the output of large language models, to training open source alternatives.
You can find this video and the entire library Vox videos on YouTube.
Why AI art struggles by hand
Subscribe to our channel and turn on notifications Don’t miss the next three episodes of this series on machine learning.
back to top ↑
