Google Bard, the company’s generative AI chatbot and ChatGPT rival, is getting an update today that the company claims will significantly enhance its capabilities. The company says Bard will now be powered by Gemini, Google’s newest and most advanced AI model, giving the chatbot more advanced reasoning, planning, understanding and other capabilities.
Gemini comes in three sizes, Ultra, Pro and Nano, allowing it to run on anything ranging from mobile devices to data centers.
The rollout of Gemini to Bard will take place over two phases. Initially, Bard will be upgraded with a specifically tuned version of Gemini Pro. Next year, Google will introduce Bard Advanced, which will give users access to the best AI model, starting with Gemini Ultra.
The version of Bard with Gemini Pro will first become available in English in more than 170 countries and territories worldwide, with more languages and countries, including the EU and U.K., soon.
Before launching to the public, Gemini Pro was run through a series of industry standard benchmarks, and in six out eight of those benchmarks, Gemini outperformed GPT-3.5, Google says. That includes better performance on MMLU, or the massive multitask language understanding tasks, which is one of the key standards for measuring large AI models. It also outperformed on GSM8K, which measures grade school math reasoning. However, as TechCrunch’s Kyle Wiggers pointed out, GPT-3.5 is over a year old, which makes this launch feel more like a catch-up rather than an outperforming.
The improvements will make Bard more capable in terms of things like understanding and summarizing content, reasoning, brainstorming, writing and planning, the company notes.
“This is the biggest single quality improvement of Bard since we’ve launched,” said Sissie Hsiao, VP and GM of Assistant and Bard at Google, when introducing the Bard upgrade in a press briefing.
Gemini Pro will first power text-based prompts in Bard to start, Hsiao said, but it will expand to multimodal support — meaning texts and images or other modalities — in the coming months.
In 2024, Bard Advanced will debut, which will be a new experience powered by Gemini’s most capable model. With Gemini Ultra, as it’s called, the AI can understand and act on different types of information, including text, images, audio, video and code, and it has multimodal reasoning capabilities. Gemini Ultra can also understand, explain and generate high-quality code in popular programming languages, Google says, in addition to understanding audio and video content. This upgrade seems to be the one to wait for, in that case.
The company says it will launch a trusted tester program for Bard Advanced before opening it up more broadly to users early next year. In addition, Google will be putting Bard Advanced through additional safety checks prior to its launch.
The update follows a number of other improvements to Bard, since its debut just eight months ago. In recent months, the AI experience has been improved with features like the ability to answer questions about YouTube videos, as well as tap into users’ Google apps, like Gmail, Docs, Drive and more, plus other Google services like Google Flights and hotels. It can also double-check its answers to help determine if the AI is “hallucinating” — that is, when it provides a response based on false information.
“Now with Gemini, we’re one step closer to bringing you the best AI collaborator in the world,” Hsiao noted. That at least seems more honest, as it’s an admission that Bard is not quite there yet.
Bard with Gemini is available today.
Comment