Google’s Bard receives major upgrade with Gemini pro as generative AI market heats up
Google has unveiled its largest and most powerful artificial intelligence (AI) model yet, Gemini 1.0, as part of a substantial upgrade to its generative AI tool Bard. Gemini, which boasts multi-modal reasoning capabilities, will be available in three different sizes – nano, pro, and ultra – catering to a range of devices from data centers to mobile phones. Initially, Bard will utilize a specifically tuned version of Gemini pro in English to enhance reasoning, planning, coding, summarizing, understanding, and interpretations. The upgrade demonstrates Google’s determination to compete with rivals Microsoft-backed Bing and ChatGPT in the thriving generative AI market. Sissie Hsiao, VP and general manager for Assistant and Bard at Google, highlighted that Gemini pro outperformed GPT 3.5, including in the vital field of massive multitask language understanding. Additionally, early next year, Google plans to launch Bard Advanced, starting with Gemini ultra, further enhancing the user experience. While currently available in English in over 170 countries and territories, more languages will be added in the coming months. Sundar Pichai, the CEO of Alphabet, emphasized that the company is only just scratching the surface of what is achievable with generative AI. Google aims to integrate Gemini into its range of products and services, including Search, Ads, Chrome, and Duet AI, and offers developers and enterprise customers access to Gemini pro for customization purposes.
Gemini ultra, expected to be released next year, is designed for complex tasks, swiftly comprehending and acting upon various types of data such as text, images, audio, video, and code. Google confirmed that extensive safety checks are being conducted, with plans to launch a trusted tester program before making Bard Advanced accessible to a wider audience. Gemini nano, on the other hand, has already been introduced on Google’s Pixel 8 Pro smartphones. Google also shared that it has been experimenting with Gemini in Search, resulting in a 40% reduction in English latency in the US and improved quality. The California-based tech giant is actively testing Gemini models across numerous tasks, including natural image, audio, and video understanding, as well as mathematical reasoning. In fact, Gemini ultra has achieved superior performance on 30 out of 32 widely used academic benchmarks. Impressively, it is the first model to outperform human experts on massive multitask language understanding, which covers subjects ranging from history and medicine to physics and law.
Google underscores its commitment to developing AI responsibly, focusing on research ambitions that benefit individuals and society, while actively engaging with experts and governments to mitigate risks. The development of Bard aligns with Google’s AI Principles, including contextual help features like the Google it button for fact-checking answers. As Google continues to build momentum in the generative AI market, it seeks to unlock the vast potential of this technology while prioritizing user safety and delivering high-quality results.
