Ever since OpenAI’s ChatGPT took the world by storm late last year, we’ve been waiting for the inevitable competition to arise. Today, OpenAI’s first significant rival hit the scene. After revealing the Gemini AI model at Google I/O 2023, Google officially launched Gemini 1.0 on Tuesday. The company calls Gemini its largest and most capable model to date, with the ability to understand and combine text, images, audio, video, and code.
Google’s Gemini model comes in three different sizes: Ultra, Pro, and Nano. Ultra is the largest model, capable of handling large and complex tasks. Pro is in the middle and can scale across a wide variety of tasks. Nano is the most efficient model and is intended for on-device tasks. From data centers to smartphones, Gemini is up for the task.
Google CEO Sundar Pichai and DeepMind CEO Demis Hassabis explained how Gemini differs from other multimodal models: “We designed Gemini to be natively multimodal, pre-trained from the start on different modalities. Then we fine-tuned it with additional multimodal data to further refine its effectiveness. This helps Gemini seamlessly understand and reason about all kinds of inputs from the ground up, far better than existing multimodal models — and its capabilities are state of the art in nearly every domain.”
They also shared data comparing Gemini to GPT-4, and Gemini frequently came out on top. In 30 of 32 widely-used academic benchmarks for LLM research and development, Gemini Ultra exceeded results from GPT-4. Gemini Ultra is also the first model to beat human experts on MMLU (massive multitask language understanding), with a score of 90%.
There are already a few ways to see Gemini in action. Starting today, Bard will use a fine-tuned version of Gemini Pro for more advanced functionality. Google says this is its biggest upgrade to Bard since the conversational AI launched. Google will also bring Gemini Nano to the Pixel 8 Pro to power features like Summarize in the Recorder app and Smart Reply in Gboard. In the future, Gemini will come to Search, Ads, Chrome, and Duet AI as well.
On December 13, developers and enterprise customers will be able to access Gemini Pro using the Gemini API in Google AI Studio or Google Cloud Vertex AI. Android developers can sign up for an early preview of AICore to start building with Gemini Nano.
If you want to learn more about Gemini, be sure to read Google’s full announcement.