The 10 Biggest Things We Learned About Google's New Gemini AI

You’ll have to wait a while before you get to try out Gemini’s most powerful capabilities, but a mid-tier version of the model is already live on Bard, the company’s chatbot. Google called this Bard’s biggest update yet. According to the company, it brings the often-overlooked product in line with the free version of ChatGPT, which runs on a model called GPT-3.5.

Google announced three different versions of the new AI. Gemini Ultra is Google’s most powerful model, pitched as a competitor to OpenAI’s GPT-4. Gemini Pro, currently running on Bard, is a mid-range model powered to beat out GPT-3.5, the baseline version of ChatGPT. Last is Gemini Nano, a more efficient model built to run on mobile devices.

Gemini Ultra is the most interesting prospect here, but only a short list of people will get their hands on it. For now, Ultra is only coming to hand-picked testers, safety experts, and major business partners. The rest of us won’t see Gemini until sometime “early next year.”

The Pixel 8 Pro is the first smartphone to get Gemini Nano. It won’t do much at first. Gemini Nano now powers the Summarize feature on Android’s Recorder app on Pixel 8 Pros. Google says the AI will also power Android’s Smart Reply feature on the Pixel 8 Pro, but only if you’re using the Google keyboard, and only in WhatsApp. The company says Gemini is coming to more messaging apps and other parts of the operating system next year.

Google is generally reserved when it takes shots at competitors, but the company didn’t mince words. According to Google, Gemini has OpenAI beat on almost every measure.

“With a score of over 90%, Gemini is the first A.I. model to outperform human experts on the industry standard benchmark MMLU,” said Eli Collins, Vice President of Product at Google DeepMind, speaking at a press conference. “It’s our largest and most capable A.I. model.” MMLU, short for Massive Multitask Language Understanding, measures AI capabilities using standard tests in a combination of 57 subjects such as math, physics, history, law, medicine, and ethics.

“Gemini’s performance also exceeds current state-of-the-art results on 30 out of 32 widely used industry benchmarks,” Collins said.

Gemini Pro is running on Bard right now, but if you want to chat with the best version of Google’s AI, you’ll have to pay. Google said it will launch a paid tier of Bard called Bard Advanced, though the company declined to share details on pricing. The premium tier will come with access to Gemini Ultra and other still unannounced features.

OpenAI already does a tidy business with its ChatGPT Plus, which costs $20 a month. Bard Advanced will open up a brand new revenue stream for Google that could be a winner in the long term.

Right now, the only modern chatbot that will speak to you is ChatGPT, and it’s mostly tight-lipped. You can also use voice chat on the ChatGPT mobile app, but there’s no “Hey ChatGPT” feature to call it up automatically.

That may change with Bard. Google says it’s adding Bard to Google Assistant sometime next year. It’s unclear exactly what that will look like, but at a Tuesday press conference, Google shared a video that featured Bard speaking to a user. When smart assistants like Siri, Alexa, and Google Assistant get a truly capable AI engine, rather than the limited pre-canned responses we get today, it will revolutionize our relationship with AI.

Google made a big deal about Gemini’s “multimodal” capabilities, “multimodal performance,” meaning it can comprehend different kinds of information such as text, images, video, audio, and more. According to the company, Google trained Gemini on a variety of mediums from the ground up, rather than taking it on after the chat features were up and running.

Google shared a video where a Gemini-powered Bard helps with a student’s physics homework starting with a photo of the assignment with handwritten questions. The AI then seamlessly transitions to written advice, complete with equations and step-by-step answers.

Two years ago, Google launched AlphaCode, a robot code writer that was the first AI smart enough to be competitive in coding conferences. Computer engineers were among the earliest adopters of tools like ChatGPT, thanks to AI’s ability to translate ideas written in human language to machine-readable code.

Along with Gemini, Google’s releasing a new version of that tool called AlphaCode 2. According to the company, it “excels at solving competitive programming problems that go beyond coding to involve complex math and theoretical computer science.”

Gemini isn’t just going to run Bard and on-device AI with Android phones; Google said the new model will be utilized across all of the company’s most important products, including Chrome, Search, Ads, and more. There’s no real timeline. Google would only say these products are getting Gemini power “in the coming months.”

One of Google’s biggest money makers is a side of the company you interact with every day, but one most people never think about. Google Cloud offers a variety of services for businesses, including data storage, data analytics and machine learning, and a set of management tools. A huge portion of the business world’s technology stack runs on Google Cloud, and Gemini is set to usher in the next era of the platform.

Along with the broader Gemini announcements, Google unveiled Cloud TPU v5p, a new version of its computer chips built for AI that companies can harness remotely through Google Cloud. Building and training AI tools requires an enormous amount of processing power, and Google is setting itself up for success with chips optimized for the task — especially because it’s building its own AI models to run Cloud TPU v5e, which will encourage more people to choose Gemini over GPT-4 and other offerings. It’s wonky stuff, but it’s a sign that Google is making huge bets on its AI business.