Google announces chatbot ‘Meena’: Key things to know

Google has announced its new chatbot named 'Meena' and the prowess that goes behind it. The firm not just revealed what 'Meena' can do but also boasted it as the world's first ever chatbot that has reached the closest to humans when it comes to 'Sensibleness and Specificity Average'. While a lot of technical information was given in the blog post by Google researchers. Here we have broken down all of that in simple terms, listing out what it means for end users and what powers the new chatbot.

'Meena' is not available for end users just yet.

Google has clearly mentioned that its new chatbot is not available for all. However, it may reach more users in the coming months. "We are evaluating the risks and benefits associated with externalizing the model checkpoint, however, and may choose to make it available in the coming months to help advance research in this area," said Google in a blog post.

Also read: Coronavirus outbreak: Google launches SOS Alert with WHO, to show helpful tips in result page

Google's 'Meena' is an open-domain chatbot.

The open-domain platform, which lets users talk and ask queries from any domain and take the conversation in any direction.

Can we classify Amazon Alexa, Google Assistant or Apple's Siri as open-domain chatbots?

No you cannot. These platforms are more of virtual assistants and offer more services than what chatbots do. Chatbots are not the same as virtual assistants.

Also read: Android phones are harder to crack than iPhones, according to a forensic detective

The power behind Google's chatbot 'Meena'?

The firm says that 'Meena' was trained with 40 billion words and 341GB of text data including social media conversations. It is based on Google's Seq2seq model, a neural network that reads words placed next to each other in a paragraph and checks if the relation between those two makes sense.

'Meena' is still not perfect.

In the blog post, Google confirmed that the chatbot has not yet achieved perfection. "While we have focused solely on sensibleness and specificity in this work, other attributes such as personality and factuality are also worth considering in subsequent works. Also, tackling safety and bias in the models is a key focus area for us, and given the challenges related to this, we are not currently releasing an external research demo," said the researchers.