Microsoft, Google and Facebook are all pursuing chatbots, which will function as virtual assistants, answering questions, responding to requests, and anticipating needs. But building functioning chatbots, which are based on artificial intelligence, is harder than it sounds. To further progress, Google open-sourced one of its natural language tools. Although Facebook hasn’t yet open-sourced it, the company introduced DeepText, a natural language engine that it is just beginning to use with its own services.

According to Wired, Google and Facebook are using “deep neural networks,” which are also used to recognize faces in photos and commands spoken into smartphones.

Google’s open-sourced SyntaxNet can understand “the grammatical logic” of a sentence using neural nets, which it does via syntactic parsing or analyzing millions of sentences. But SyntaxNet has its limits, and the company continues the work to enable it to “understand the complete meaning of a conversation.”

Facebook’s DeepText is based more on math, which, says Chris Nicholson, “can make for a more flexible system.” Nicholson is founder of deep learning startup Skymind.

Facebook director of engineering Hussein Mehanna reports DeepText already works with 20 languages, and that the company is testing it as a way to power chatbots inside Facebook Messenger. Facebook also has the data — 400,000 new posts every minute — to learn from natural language.

At Google, machine language pioneer Ray Kurzweil, who was hired in 2012 to work on natural language recognition, says he and his team are building a chatbot, dubbed Danielle after a character in one of his books, to be released sometime this year, says The Verge.

Kurzweil notes that “having meaningful conversation with artificial intelligence” is his goal, although he believes that bots won’t pass the Turing test (that is, to be indistinguishable from humans) until 2029. Eventually, he says, anyone will be able to create their own chatbot by feeding it a “large sample of writing” such as a blog.

In China, says TechCrunch, WeChat is an ubiquitous app that’s used for pretty much all digital services. “Tencent’s messaging app might even be giving China an economic advantage over the West,” it notes since, in the U.S., people use multiple messengers and email whereas in China, it’s WeChat for everything, all the time.

WeChat, says product manager Dan Grover, is “an effective solution for a broken OS,” exemplified by “meaningless notifications, a lack of baked-in QR code scanners, bloated apps that hog memory, contacts that are disjointed from social graphs, cumbersome authentication systems and an absence of universal payments solution.”

Although chatbots are a nice addition, “nothing in the U.S. has been able to fix any more than one or two” of the OS problems Grover lists. By comparison, WeChat is a “tight ecosystem… a connected web of services.” Chatbots might one day be part of the ecosystem, says TechCrunch, but never all of it.