Technology
Wednesday, December 6th, 2023 3:40 pm EDT
Key Points
- Introduction of Gemini AI Model Suite: Google has launched Gemini, its largest and most advanced artificial intelligence (AI) model, addressing increasing pressure to elucidate its AI monetization strategy. The Gemini suite comprises three models: Gemini Ultra, the most extensive and capable; Gemini Pro, versatile across various tasks; and Gemini Nano, tailored for specific tasks and mobile devices.
- Application and Features of Gemini: Initially, Google plans to license Gemini to customers through Google Cloud for integration into their applications. Developers and enterprise customers can access Gemini Pro starting from December 13 via the Gemini API in Google AI Studio or Google Cloud Vertex AI. Android developers will also be able to build with Gemini Nano. The AI model will power Google products such as the Bard chatbot and Search Generative Experience, offering conversational-style responses to search queries.
- Gemini’s Capabilities, Bard Integration, and Technical Aspects: Gemini Ultra, part of the Gemini suite, is highlighted as the first model to outperform human experts in massive multitask language understanding (MMLU), encompassing subjects like math, physics, history, law, medicine, and ethics. Sundar Pichai, CEO of Alphabet Inc., emphasizes Gemini’s multimodal capabilities, allowing it to understand and combine different types of information, including text, code, audio, image, and video. Bard, Google’s chatbot, will use Gemini Pro for advanced reasoning, planning, and understanding. Google executives mentioned the efficiency of Gemini Ultra, stating it is significantly cheaper to serve and emphasizes its enhanced efficiency in model training. While a technical white paper detailing the model will be released, Google refrains from disclosing the perimeter count.
Google has unveiled its latest and most advanced artificial intelligence (AI) model, Gemini, amid growing pressure for the tech giant to articulate its AI monetization strategy. The model is expansive, featuring three categories: Gemini Ultra, the largest and most capable; Gemini Pro, versatile across various tasks; and Gemini Nano, designed for specific tasks and mobile devices. Initially, Google plans to license Gemini to customers via Google Cloud for use in their applications.
Gemini Ultra, the flagship model, reportedly surpasses human experts in massive multitask language understanding (MMLU), covering 57 subjects like math, physics, history, law, medicine, and ethics. This model, representing Google’s commitment to multimodal capabilities, claims to comprehend nuance and reasoning in complex subjects. Sundar Pichai, CEO of Alphabet Inc., stated that Gemini is built to seamlessly understand and combine different information types, including text, code, audio, image, and video.
Gemini will be available for use by developers and enterprise customers through Google Cloud from December 13. Gemini Pro can be accessed via the Gemini API in Google AI Studio or Google Cloud Vertex AI, while Android developers will have access to Gemini Nano. Google intends to employ Gemini in its products, such as the Bard chatbot and Search Generative Experience, the latter aiming to provide conversational-style responses to search queries.
The announcement also highlighted that Google’s chatbot Bard will now utilize Gemini Pro for advanced reasoning, planning, and understanding. An enhanced version, “Bard Advanced,” leveraging Gemini Ultra, is set to launch early next year.
Addressing questions about Gemini’s novel capabilities and comparisons with competitors like GPT-4, Google executives remained elusive. Sissie Hsiao, Google’s general manager for Bard, stated that the focus is on creating a good experience, with no current plans for monetization details.
Despite a delayed launch, Google emphasizes Gemini’s extensive testing and safety evaluations, asserting it as the most highly tested AI model to date. Gemini Ultra is highlighted for not only being more capable but also more cost-efficient to serve. Google plans to release a technical white paper for Gemini but will not disclose the perimeter count.
In addition to Gemini, Google introduced the next-generation tensor processing unit, TPU v5p chip, for training AI models. While details on performance compared to market leader Nvidia were not provided, Google claimed improved performance for the price compared to the TPU v4 announced in 2021.
The AI-related announcements come as investors seek clarity on Google’s strategy to turn AI into a profitable venture. The company had previously launched the “Search Generative Experience” experiment in August, offering a more conversational search experience. However, specifics on the public launch remain vague, with executives mentioning Gemini’s incorporation into it within the next year.
Sundar Pichai expressed excitement about Gemini, labeling it one of the company’s most significant science and engineering efforts, unlocking opportunities for people worldwide.
For the full original article on CNBC, please click here: https://www.cnbc.com/2023/12/06/google-launches-its-largest-and-most-capable-ai-model-gemini.html
