People who have the Inbox email program on their iPhones or Android devices will soon have a new option when it comes to replying to emails. Instead of coming up with their own responses on their mobile devices, they’ll get to choose between three options created by a neural network built by Google researchers.
Google claims it has built an AI that can read incoming emails, understand them, and generate a short, appropriate response that the recipient can then edit or send with just a click.
In the case of Smart Reply, what Google has done is combine several systems to build a neural network that can read your email, parse what the words in the email mean, and then not only generate a response, but generate three different responses. This is more than just building out rules for common words that fall in an email. This is truly teaching a computer to understand the text of an email. It uses the type of neural networks found in natural language processing to understand what a person means and also generate a reply.
Another bizarre feature of our early prototype was its propensity to respond with “I love you” to seemingly anything. As adorable as this sounds, it wasn’t really what we were hoping for. Some analysis revealed that the system was doing exactly what we’d trained it to do, generate likely responses—and it turns out that responses like “Thanks”, “Sounds good”, and “I love you” are super common—so the system would lean on them as a safe bet if it was unsure. Normalizing the likelihood of a candidate reply by some measure of that response’s prior probability forced the model to predict responses that were not just highly likely, but also had high affinity to the original message. This made for a less lovey, but far more useful, email assistant.