Google has filed a patent for a personalized artificial intelligence (AI)-powered chatbot that can answer — and respond to — phone calls on your behalf.
The U.S. Patent and Trademark Office published the patent filing on Tuesday (Feb. 11) for a chatbot that can automatically answer a call, understand what the caller wants and respond to the request. This is the first time the patent has been revealed to the public.
The AI chatbot can be embedded in smartphones, laptops, desktop computers, tablets, wearable computing devices, car systems and others, according to the patent filing.
Google called the AI system “personalized chatbots that can answer calls on behalf of the users and dynamically adapt a level of interaction required by the users.”
Moreover, the AI chatbot makes the decision whether to completely handle the call itself, loop you in on a decision to make, or not interfere in a call.
How the Personalized Chatbot Would Work
Google said the chatbot can help minimize distractions caused by routine phone calls in everyday life – whether from the delivery person, restaurant reservation confirmations and the like.
For example, if a delivery company calls to find out where to place a package, the chatbot can tell the caller “outside the door” based on its knowledge of your preferences.
If a more involved answer is needed — for example, the delivery company is calling to tell you the package is damaged and asks whether you want to reorder the product or get a refund — the chatbot will understand the question and send you a choice of “re-order” or “refund” buttons to press or audibly ask you.
Google said the chatbot will tailor its questions based on who is calling. It could ask you to decide “yes” or “no” on a matter if this would suffice.
The chatbot can also be set to not interfere in certain priority phone calls — say from the day care center — and instead send them directly to you.
On-Device AI Model
The chatbot taps into a trend of edge AI, where the AI model is small enough to reside entirely in the mobile device and the computer chip is sufficiently powerful enough to do the processing.
Google said it uses an on-device machine learning model engine that can include automatic speech recognition (ASR), a natural language understanding engine (NLU), a fulfillment engine and a text-to-speech engine.
However, it can also access the cloud if necessary, although the model being on-device means less lag time and more privacy.
The chatbot can access personal data on the device, such as text messages, email, profiles and other information with your approval.
Google plans to develop the AI chatbot but will also allow third parties to do so.