Baidu’s Ernie AI Gets Improved Text-to-Image and App Builder

Ernie, the foundation model for Baidu’s generative AI, has been updated with iRAG technology to mitigate visual hallucinations and a no-code tool called Miaoda that creates apps using natural language. The company behind China’s largest search engine says Ernie now handles 1.5 billion daily user queries, up from 50 million circa its March 2023 launch (a 30x increase). Baidu also debuted Ernie-powered smart glasses from its Xiaodu Technology hardware unit. The Xiaodu AI Glasses features built-in voice activation and cameras for taking photos and video. The news was shared at this week’s Baidu World 2024 in Shanghai.

Speaking at the conference, Baidu CEO Robin Li emphasized Ernie’s rapid adoption, which was at 200 million Chinese users in May.

Ernie is not available outside of China, where the Shanghai-based Baidu is among the “tech firms shifting their focus to the commercialization of large language model (LLM) applications after nearly two years of heavy investment in research and development in models that they tout as alternatives to OpenAI’s GPT,” reports Reuters.

At Baidu World, Li introduced iRAG, “a text-to-image technology that leverages Baidu’s search capabilities to address the ‘hallucination’ issue,” Reuters says, “referring to the generation of images that deviate from the input text or contain non-existent elements.”

Li also unveiled Miaoda, a no-code tool that lets Ernie users build apps by describing them in natural language. “Miaoda provides no-code programming, multi-agent collaboration, and multi-tool invocation,” Baidu explains in a news post, explaining that codeless programming “allows anyone to generate code without writing a single line, lowering barriers to AI development and making it accessible to all.”

Multi-agent collaboration lets Ernie “coordinate and manage different agents effectively, while its multi-tool invocation taps into Ernie’s tool invocation abilities, extensively utilizing web search, iRAG, maps API, and other tools for a seamless workflow.”

“Baidu isn’t aiming to launch a ‘super app,’” Li said from the Baidu World stage, in reference to competitor ByteDance. Also known as an “everything app,” these multipurpose applications handle everything from digital wallets to social media and creative uses.“Instead, we aim to help more people and businesses create millions of ‘super useful’ applications,” Li said.

Xiaodu, which focuses on consumer hardware, is expected to begin shipping its AI smart glasses next year. Engadget writes that “they could become the Chinese consumers’ alternative to Meta’s and Snap’s devices.” Though Xiaodu hasn’t announced pricing, the Meta Ray-Ban smart glasses retail for $300.

Baidu rushed Ernie to market in early 2023 in attempt to beat competitors to the punch with consumer AI.

No Comments Yet

You can be the first to comment!

Leave a comment

You must be logged in to post a comment.