Gemini Ultra marks a leap forward for Google in multimodal AI technology, setting a new benchmark for AI language and coding capabilities.
On Wednesday, Google launched Gemini, its most advanced artificial intelligence system to date, marking a major milestone in its development of artificial intelligence.
Described by Google CEO Sundar Pichai as “our largest and most powerful AI model yet,” Gemini is designed to understand and act seamlessly across text, images, audio, video, and other information modalities.
“This new era of modeling represents one of the biggest scientific and engineering efforts we’ve ever undertaken as a company," Pichai wrote in the announcement. "I’m incredibly excited about the future and the opportunities Gemini will bring to people around the world.”
The company said the new system, optimized at three scales, significantly enhances Google’s AI products and surpasses previous benchmarks. According to Google’s testing, Gemini scored higher than any other model on more than 30 leading AI benchmarks.
Most notably, the largest version, Gemini Ultra, became the first AI system to surpass average human performance on a language test covering more than 50 academic subjects. The statement continued;
“With a score of 90.0%, Gemini Ultra is the first model to outperform human experts on MMLU (Massive Multi-Task Language Understanding), which combines 57 subjects including mathematics, physics, history, law, medicine, and ethics to test world knowledge and problem-solving abilities.”
Gemini has also demonstrated new capabilities in computer programming. Google used a specialized version of Gemini to create an advanced code generation system called AlphaCode 2 that it said was “excellent at solving competitive programming problems.”
When evaluated in programming competitions, AlphaCode 2 solved nearly twice as many problems as its predecessor, and Google estimates that it will outperform 85% of human participants.
Google is releasing Gemini on several of its products right away, including the Bard conversational AI tool it launched earlier this year and its Pixel smartphones.
"Starting today, Bard will use a fine-tuned version of Gemini Pro for more advanced reasoning, planning, understanding, and more. This is the biggest upgrade to Bard since its launch," Google said.
The Pixel 8 Pro also becomes the first phone to support Gemini Nano, powering new on-device features like text summaries in apps.
Some developers and partners will be able to experience Gemini Ultra later this year, and then it will be widely available through the Google Developer Platform in 2023. Google said it is currently completing a security review of the model. #谷歌 #Gemini
