World’s first MLLM-judge based on Google Gemini launched

Post Thumbnail

Patronus AI has announced the launch of the industry’s first MLLM-as-a-Judge (multimodal language model as a judge) technology, designed to revolutionize the evaluation of AI systems that work with images and text. Etsy, the largest marketplace for vintage and handmade goods, is already implementing this innovative technology to verify the accuracy of product image captions.

In an exclusive interview with VentureBeat, Patronus AI co-founder Anand Kannapan shared details of the collaboration with Etsy: “We are incredibly excited to announce that Etsy has become one of our first clients. Their marketplace contains hundreds of millions of handmade and vintage items from around the world. Etsy’s AI team sought to use generative artificial intelligence to automatically create image captions, ensuring their correctness when scaling across their entire global user base.”

A key feature of the new Judge-Image technology is the use of Google’s Gemini model instead of OpenAI’s GPT-4V. This decision was made after a thorough comparative analysis of available alternatives. “We noticed that GPT-4V demonstrated a more pronounced tendency toward egocentricity, whereas Gemini showed itself to be less biased and demonstrated a fairer approach to evaluating various input and output pairs,” explained Kannapan. “This is confirmed by the even distribution of scores across different sources.”

The new technology aims to address a critically important problem in artificial intelligence – identifying and minimizing hallucinations and reliability issues in multimodal AI applications. For Etsy, whose business model is built on trust between sellers and buyers, the accuracy of product descriptions is a key success factor.

The launch of Judge-Image marks an important step in the development of verification technologies for AI-generated content. As more companies incorporate generative artificial intelligence into their business processes, the need for reliable evaluation and quality control tools becomes increasingly relevant. The successful implementation of the technology on the Etsy platform could become an exemplary case for other e-commerce companies and technology platforms.

AIvengo > Reviews > World’s first MLLM-judge based on Google Gemini launched
Почитать из последнего
UBTech will send Walker S2 robots to serve on China's border for $37 million
Chinese company UBTech won a contract for $37 million. And will send humanoid robots Walker S2 to serve on China's border with Vietnam. South China Morning Post reports that the robots will interact with tourists and staff, perform logistics operations, inspect cargo and patrol the area. And characteristically — they can independently change their battery.
Anthropic accidentally revealed an internal document about Claude's "soul"
Anthropic accidentally revealed the "soul" of artificial intelligence to a user. And this is not a metaphor. This is a quite specific internal document.
Jensen Huang ordered Nvidia employees to use AI everywhere
Jensen Huang announced total mobilization under the banner of artificial intelligence inside Nvidia. And this is no longer a recommendation. This is a requirement.
AI chatbots generate content that exacerbates eating disorders
A joint study by Stanford University and the Center for Democracy and Technology showed a disturbing picture. Chatbots with artificial intelligence pose a serious risk to people with eating disorders. Scientists warn that neural networks hand out harmful advice about diets. They suggest ways to hide the disorder and generate "inspiring weight loss content" that worsens the problem.
OpenAGI released the Lux model that overtakes Google and OpenAI
Startup OpenAGI released the Lux model for computer control and claims this is a breakthrough. According to benchmarks, the model overtakes analogues from Google, OpenAI and Anthropic by a whole generation. Moreover, it works faster. About 1 second per step instead of 3 seconds for competitors. And 10 times cheaper in cost per processing 1 token.