Google’s Meena, AI assistants about to get a lot smarter
Google Artificial Intelligence assistants showed that a chatbot based on a gigantic neural network and huge amounts of data can hold coherent conversations, maintaining context over multiple turns and conversing on a topic and the name of the chatbot is called meena.
This model is based on sequence-to-sequence on transformer architecture. These are widely used tools, but this is the first time anyone’s trained such a model at this scale. It is an open domain system that means, it can talk about anything the user wants to.
Public relations risk in AI, Google tells us this is related to safety. They talk about “public domain social media conversations. It has released this publicly yet. In the research phase, the authors used an additional filter to remove “sensitive or toxic response.
Meena’s model in the future, but clearly they don’t want an online demo on a Google domain or in any way affiliated with Google branding, said by Google
The researches says that “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.
Chatbot require a lot of ingenuity to take these results and translate them into practical improvements for real-world AI assistants.
