Today Google unveiled Gemini, which is its new, “largest and most capable” AI model. It’s built from the beginning to be multimodal, and so it can generalize and understand different types of information – text, images, audio, video, and code – at the same time. This lets it better parse nuances and makes it better at answering questions relating to complicated topics. Thus, it’s especially good at explaining reasoning in complex subjects like math and physics.
It comes in three ‘sizes’ – Ultra, Pro, and Nano. Ultra is “the largest and most capable model for highly complex tasks”, Pro is the “best model for scaling across a wide range of tasks”, while Nano is the “most efficient model for on-device tasks”.
Gemini Ultra is able to understand, explain, and generate high quality code in Python, Java, C++, and Go. It can work across languages and reason about complex information. It excels in several coding benchmarks, including HumanEval and Natural2Code, Google’s internal held-out dataset, which uses author-generated sources instead of web-based information.
Google’s Bard AI is now using a “fine-tuned version” of Gemini Pro, whatever that means. This should give it “more advanced reasoning, planning, understanding, and more”, Google says. Will it make it better than ChatGPT? Your guess is as good as ours, but Google would surely want you to think so.
In fact, it boasts that Gemini Pro outperformed GPT-3.5 in six out of eight benchmarks it ran. What’s more, Google ran some blind evaluations with third-party raters and found that “Bard is now the most preferred free chatbot compared to leading alternatives”. No, we don’t know what “alternatives” Google is talking about either, because it hasn’t named them.
Anyway, you can try Bard with Gemini Pro today for text-based prompts in English in 170 territories but not Europe (which is coming soon), while early next year Gemini Ultra will be brought to “a new Bard Advanced experience”.
But wait, there’s more! The Pixel 8 Pro is, surprise, the first smartphone “engineered for Gemini Nano”, again, whatever that means. And yes you read that right: just the Pixel 8 Pro, not the Pixel 8 too. The Pro’s Summarize in Recorder and Smart Reply in Gboard features are apparently using Gemini Nano already.
The company is also “starting to experiment with Gemini in Search”, where it is making the experience faster. And, “in the coming months”, Gemini will power features in more of Google’s products and services, “like Ads, Chrome, and Duet AI”.
