Have you heard of the conversational agent, or chatbot, called Meena? In…

Have you heard of the conversational agent, or chatbot, called Meena? At the end of January, Google AI introduced it. It is a robust model with 2.6 billion parameters. The model's architecture is based on 13 decoding blocks of the Evolved Transformer seq2seq. The model is trained on 341 GB of text, primarily conversations from social networks (making it nearly twice the size of GPT-2 and trained on 8.5 times more data). In SSA (Sensibleness and Specificity Average) tests and with a low perplexity rate, it clearly outperforms competing chatbots such as Mitsuku, Cleverbot, DialoGPT, and Xiaolce.

The problem with older chatbots is that they excel only in a narrowly defined area, and if you steer the conversation elsewhere, they often struggle. For example, for language learning or interactive games, a bot with a wide range of conversational topics would be beneficial. Today's chatbots also frequently produce nonsense. They generate responses that contradict what has already been said, lack basic knowledge of the world, and common sense. All too often, they tend to respond with phrases like "I don't know." I can confirm this from personal experience. When I created my chatbot about a year ago as a final project at the Moscow School of Artificial Intelligence (NRUHSE), it too often replied with "I don't know" :)

Future research on Meena will aim to improve attributes such as personality and realism. The model has not yet been made available to the public, although discussions about it have been ongoing since the end of January.

Blog: https://ai.googleblog.com/2020/01/towards-conversational-agent-that-can.html
Paper: https://arxiv.org/abs/2001.09977
Examples: https://github.com/google-research/google-research/tree/master/meena

Originally published on Facebook — link to post