Google has officially entered the next era of artificial intelligence with the unveiling of Gemini, its latest and most powerful large language model. The announcement was made at the AI Summit in New York 2023, where Google CEO Sundar Pichai highlighted the model's capabilities and its immediate applications, particularly powering the Bard chatbot and enhancing various Google products.
Gemini's Three Sizes and Applications
Gemini comes in three sizes - Ultra, Pro, and Nano - enabling it to operate seamlessly across a wide range of devices, from mobile phones to data centers. Google has wasted no time in implementing Gemini, with the initial rollout powering the Bard chatbot through Gemini Pro. The company claims this will significantly enhance Bard's reasoning and understanding abilities, marking the start of what Google is calling the "Gemini era."
In the coming months, Gemini is set to play a crucial role in Google's Pixel 8 Pro handset, providing users with access to new features such as Summarize in the Recorder app and a Smart Reply developer preview in Gboard. Moreover, Gemini is set to make its mark on Google Ads, Chrome, and Duet AI, indicating a comprehensive integration across multiple platforms.
Gemini's Impact on Search and Beyond
Google has been experimenting with Gemini in its search algorithms, aiming to improve the Search Generative Experience (SGE) and make the process faster for users. While there's no explicit mention of Gemini powering Makersuite features in the blog post, a leaked October report suggested potential benefits for Google's web development platform.
Developer Access and Enterprise Rollout
Starting December 13, developers and enterprise customers will have access to Gemini Pro through the Gemini API in Vertex AI or Google AI Studio. The broader rollout of Gemini Ultra is planned for early 2024, with select groups gaining access before it becomes available to enterprise customers.
For Android developers looking to build Gemini-powered apps on-device, an early preview of Gemini Nano is available via Android AICore.
Competing with ChatGPT and Microsoft's AI Copilot
Gemini was positioned as a rival to OpenAI's ChatGPT, with Google keen to release the initial version as soon as possible. Despite initial plans for private events to showcase Gemini, CEO Sundar Pichai scrapped these due to concerns over the model's handling of non-English queries. Google is eager to compete not only with ChatGPT but also with Microsoft's AI Copilot, which has gained significant traction in powering productivity software.
New TPU Chips and AI Hypercomputer
Google has been training Gemini on its in-house TPU chips, unveiling new AI accelerators, Cloud TPU v5p, that train large language models 2.8 times faster than the previous generation. The company also introduced an AI Hypercomputer, a supercomputer architecture designed to meet the exponential demands of generative AI. This integrated system of AI-optimized hardware and software, powered by the new TPU v5ps, provides a secure computing environment for modern AI workloads.
Access to Cloud TPU v5p or the new AI Hypercomputer requires enterprises to contact their respective Google Cloud account manager.
In conclusion, Google's Gemini is set to redefine the landscape of AI applications, powering a wide range of products and services with its cutting-edge capabilities. As enterprises and developers gain access to this next-generation language model, the AI revolution is poised to reach new heights.
