Now AI bot learns to fake conversations human way

Dec 27, 2017, 6:41 AM EST
(Source: James Royal-Lawson/flickr)
(Source: James Royal-Lawson/flickr)

Artificial intelligence is being schooled to pick another cheeky human attribute but this time, to make man-machine chats smoother. A team of researchers in Japan are training AI chatbots to learn without asking pesky questions, which make engagements monotonous and frustrating at times.

The new method prompts an ignorant AI chatbot to fake its way through conversations instead of repeatedly asking direct questions, a technique that the team calls “lexical acquisition through implicit confirmation,” writes Inverse.

This is a trick where human respondent is drawn into confirming or denying the AI’s instinct, which saves the bot from appearing as an annoying fellow who learns only by asking.