San Francisco, Aug 6 : Meta’s AI research labs have developed a state-of-the art chatbot and Meta allows users of the public interact with the system in order to gather comments on the capabilities of its system.According to The Verge, the bot is dubbed BlenderBot 3 and can be accessible via the internet.
However, at present it appears that only those in the US are able to access it.
“We are committed to releasing publicly all of the data we’ve collected in the demonstration in the hope to improve the quality of conversational AI.” Kurt Shuster, an engineer at research company Meta who helped develop BlenderBot 3, was quoted as saying.
BlenderBot 3 can engage in general chat, according to Meta however, it can also respond to the kinds of questions you’d like to ask an assistant in a digital format, “from talking about healthy food recipes to finding kid-friendly facilities in the city” according to the report.
The bot is a concept and is based on Meta’s previous work with what’s called large language models, or LLMS — a powerful but flawed software for creating text of which OpenAI’s GPT-3 is the most well recognized instance.
As with the majority of LLMs, BlenderBot is initially trained on massive text databases which it analyzes for statistical patterns to create language.
These systems have proven extremely adaptable and have been used for a variety of applications, from creating programming code for developers to assisting authors in writing their next novel.
But, they have serious flaws.
They reproduce the biases in the training data and frequently invent answers to user’s questions (a major issue for them if they’re going to function as digital assistants).
vc/
