Google has launched Gemini, a revolutionary AI model designed to work seamlessly with text, images, and video. Positioned as one of Google’s most significant algorithms since PageRank, Gemini aims to redefine the landscape of AI technology. The initial version is integrated into Google’s Bard chatbot for the English language setting, with broader availability for developers through Google Cloud’s API starting December 13.

More Than Just a Chatbot, Google’s New AI Aims to Redefine the Landscape

Gemini stands out as a “natively multimodal” model, distinguishing itself by being trained on images, video, audio, and text. Eli Collins, VP of Product for Google DeepMind, describes Gemini as the “largest and most capable” and highlights its generality. The model comes in three versions: Ultra (largest), Nano (smaller and efficient), and Pro (medium size and capabilities).

Google’s Bard chatbot, similar to ChatGPT, now leverages Gemini Pro, enhancing its capabilities for advanced reasoning and planning. The model will power suggested messaging replies on Pixel 8 smartphones. Gemini set to integrate into various Google products, including generative search, ads, and Chrome, in the upcoming months. Most potent version, Gemini Ultra, scheduled for a 2024 release, subject to extensive trust and safety checks.

Sissie Hsiao, VP at Google and GM for Bard, emphasizes that Gemini’s multimodal capabilities have significantly improved Bard’s skills, enabling better content summarization, brainstorming, writing, and planning. Gemini’s integration into Bard represents a pivotal moment for both technologies.

Gemini Pro outperforms its predecessor, GPT-3.5, on six out of eight commonly used benchmarks for evaluating AI software intelligence. The upcoming Gemini Ultra boasts a remarkable 90 percent score on the Massive Multitask Language Understanding (MMLU) benchmark, surpassing other models, including GPT-4. Google asserts Gemini’s state-of-the-art performance across 30 out of 32 widely used benchmarks.

The AI Arms Race Heats Up: Gemini vs. ChatGPT and Beyond

Gemini’s launch is part of an ongoing competition in the AI space, marked by OpenAI’s ChatGPT and subsequent advancements. Oren Etzioni, a professor emeritus at the University of Washington, characterizes this environment as a “tit-for-tat arms race,” anticipating further developments with future models like GPT-5.

Google acknowledges the increased responsibility associated with Gemini’s enhanced capabilities. The model underwent rigorous safety testing, including collaboration with external researchers to “red-team” the model, identifying potential weaknesses. Google’s commitment to quality and safety checking aligns with the model’s broader influence.

Gemini represents a pivotal advancement in Google’s AI endeavors, embracing multimodal capabilities to deliver a more versatile and powerful AI model. As it permeates various Google products, Gemini sets the stage for an exciting era in AI development, inviting developers and users to explore its diverse applications.