What does a neural network do?

A neural network chatbot from Google in test mode has started working at Wendy’s fast food chain. Its main goal is to reduce long queues in restaurants by transferring order data. Wendy’s executive director says it’s no different than a regular employee talking to a chatbot during a customer’s ordering process. Moreover, the recipient may not even understand who they are talking to.

• To work in fast food, AI uses a natural language model from Google that examines the entire list of items on the menu.
• The chatbot even knows how to upsell by offering to buy it with a larger serving or sauce.
• The neural network can recognize the abbreviations and slang used by customers when placing an order, but sometimes there are difficulties with this: for example, a bot offers a person a Frosties shake, but on the final order, the customer can simply search for the product. without specifying milkshakes and neural net should understand what is said about Frosties in particular.

Another problem is noise, as outside customers can be difficult to hear due to nearby road noise, music or passengers in the backseat of a car – difficult even for a human and the machine may not be able to distinguish a customer’s voice. and an external source.

At this stage, the technology collects only order-related data and passes it on to a real employee who processes everything. However, it is already successfully replacing cashiers who accept orders.

If the AI ​​performs well enough, it is likely to replace some restaurant staff in the future. It is scheduled to launch in full at one of the US institutions in June.