When Facebook launched chat bots for Messenger earlier this year, we knew the next big thing in communication was now. But then came millions of bots bugging you with their unworthy AI and headless talks.
Well, Facebook has (finally!) seen that and is now making the soul of these artificial intelligence-powered bots open for anyone to employ and study. The development comes as part of the Facebook AI Research (FAIR) lab’s mission to help researchers and engineers by making its work available generally.
Facebook’s new program makes training AI bots easier and efficient. This not only boosts the third-party offerings, but also encourages developers to focus on building for Facebook’s platform first over other rival environments.
The library of code is called fastText and can be downloaded by anyone from GitHub. The only requirement the social giant has mentioned is that you will need a compiler with “good C++11 support.” As the name suggests, fastText mainly focuses on speed and efficiency. According to its maker, the platform can train models “on more than 1 billion words in less than 10 minutes using a standard multicore CPU.” That’s a huge claim to make, but most developers have been convinced until now.
According to FAIR (really like the name here), the new system can perform days’ worth training in a matter of seconds or even less.
Talking about the whole development, FAIR said,
“Ultimately, we hope that fastText will help us all design better applications and further advance the research in language understanding.”
The basic working of fastText completely relies on basic language functions like classifying words and sentences. Following this, the system generates libraries that programs can reference when executing tasks. In fact, fastText is so refined it can interpret requests based on gender, state and even number. When hit with a query, fastText can look into the fastText-generated document and understand that the user is asking for specific information.