Google Upgrades Bard: Introducing Gemini Model, Performance Surpasses Across the Board

OpenAI's ChatGPT has become a global phenomenon and is one of the rapidly growing consumer products, while Google's Bard seems to pale in comparison. This chatbot continues to gain new features, including accessing data from other Google products, but its responses and information rarely compete with ChatGPT and other robots using GPT-3 and GPT-4.

However, Bard's situation may have just become more compelling: as of now, for English users in 170 countries, Bard is now powered by Google's new Gemini model, which Google claims is on par with OpenAI's technology in multiple aspects, and even surpasses it. Google states that Gemini will be expanded to more languages and countries "in the near future."

Bard is now running on Gemini Pro, which is the middle-tier of the Gemini series. Ultra is the largest and slowest but most capable version, while Nano is small and fast, suitable for tasks on devices, and Pro is in between. It aims to be the "just right" version of the model: both fast and efficient, and as capable as possible.

Sissie Hsiao, responsible for Bard and Assistant, said at the press conference that Gemini represents the "biggest and best upgrade" for Bard so far. It should significantly improve everything Bard has been doing: summarizing, brainstorming, writing, and more. Google CEO Sundar Pichai said in his testing, he didn't find any amazing new features, but there were overall improvements. "I think people will find this product much better," he said. "It understands their intent better and answers better. More accurate, higher quality. If you're trying to code, it does better!"

Currently, Bard is still just a chatbot: you input text, and it replies with text. But a brand new version of Bard will be launched soon. Next year, Google plans to release a preview version of "Bard Advanced," powered by the most powerful and capable Google's new large-scale language model, Gemini Ultra. Gemini Ultra is also a multimodal version of the model, which means it can accept and create images, audio, and video, not just text.

Demis Hassabis, head of Google DeepMind, said that non-text interaction is a highlight of Gemini. "We built it as a fundamentally multimodal model from the start," he said. "That's one of its new capabilities... the seamless integration and reasoning it can do across multimodal." Google's demo includes YouTuber Mark Rober using Bard to create a perfect paper airplane - including taking photos of his design for AI-provided feedback - and parents uploading photos of their children's homework to help point out their math mistakes.

However, these are currently just demonstration and promotional videos. Pichai said he believes this release is both a significant moment for Bard and the beginning of the Gemini era. But if Google's benchmark tests are correct, the new model may have already made Bard as excellent a chatbot as ChatGPT, which is already quite an achievement.