Facebook is open-sourcing its AI bot-building research

The biggest appeal of fastText appears to be speed and efficiency. According to Facebook, fastText is, as its name suggests, much quicker than other learning methods, and can train models “on more than 1 billion words in less than 10 minutes using a standard multicore CPU.” In fact, FAIR claims that, compared to deep learning models, fastText can cut training delays from several days to a few seconds. fastText focuses on classifying words and sentences, and produces libraries that programs can reference when executing tasks. For example, fastText can learn that the words “boy,” “girl,” “man” and “woman” refer to specific gendered nouns and store those values in a document. Then, when an AI program, like a bot, is interpreting a request, such as “Where my girls at,” it can…


Link to Full Article: Facebook is open-sourcing its AI bot-building research