Google’s latest “Meena” chatbot, will talk to you about anything on earth just like any other human being.
Artificial intelligence voice assistants can help you with your simple day-to-day job but you just can’t have a proper conversation with them because they don’t understand a lot of it. Meena is an end-to-end qualified neural conversational model to evaluate scientists using a new human assessment metric called the Snsibleness and Specificity Average (SSA) for open-domain chatbots. Google says the 2.6 billion-parameter conversational neural network can talk better with people than any AI generator system. The team equipped the model with 341 GB of text data including a chat on social media.
Meena has an evolved single transformer encoder block and 13 evolved decoder blocks. While encoder blocks help you understand the conversation context and decoder allows you to form an answer. Meena has 1.7x higher model power, and 8.5x more data has been educated.
It’s hard to have an open-ended discussion that encompasses a wide variety of topics and most chatbots can not. Work on the open-domain dialog explores a complementary approach that aims to create a chatbot that is not specialized but can still chat about almost everything a user wants. Many chatbots trick you into concealing the fact they don’t know what you’re doing.
For now, the Google Meena model isn’t available, much more work needs to be done.