I'm a Forbes technology reporter and senior writer based in San Francisco. Send me tips at [email protected] or [email protected].
Google DeepMind unveiled the latest version of its large language model, called Gemini 1.5, on Thursday. This is the technology underpinning its chatbot rival to OpenAI’s ChatGPT.
Google said its new model has made big advances in context, speed and efficiency. For example, Gemini 1.5 Pro, the mid-tier version of the model, can run up to 1 million “tokens” – the building block units of generative AI models – though this is currently available only to a “limited group” of developers and early testers in a private preview. By comparison, GPT-4 Turbo, OpenAI’s high-end model, has a 128,000 token limit. Google said its mid-tier 1.5 Pro model performs on par with Gemini 1.
In demo videos, researchers uploaded a transcript of the NASA moon landing and asked the model to generate funny quotes from it. Then the researcher uploaded a simple line drawing of a foot taking a step, and asked the model to pinpoint the moment in the transcript that the drawing refers to. Gemini was able to understand the drawing and point the researcher to Neil Armstrong’s famous “One small step for man, one giant leap for mankind,” quote.
The rollout for now is just a “research release,” available only to software developers and Google’s cloud customers, and not the general public. The release strategy is in contrast to rivals like Meta, which open sourced its LLaMA 2 model last year and made it widely available for use with little restriction. The rollout also differs from Google’s original launch of Gemini 1.0, which was immediately available to consumers.
“It sort of makes sense to see what creative minds—someone who is more of the developer mindset—can do with the model,” said Oriol Vinyals, vice president of research at Google DeepMind, said during a briefing with reporters on Wednesday. He added that developers will also help the company to “understand how this model will matter to users, ultimately.”The new product release follows several changes Google has recently made to its AI strategy.
Norge Siste Nytt, Norge Overskrifter
Similar News:Du kan også lese nyheter som ligner på denne som vi har samlet inn fra andre nyhetskilder.
Google renames Bard to 'Gemini' and launches Google One AI Premium plan with advanced accessJohanna 'Jojo the Techie' is a skilled mobile technology expert with over 15 years of hands-on experience, specializing in the Google ecosystem and Pixel devices. Known for her user-friendly approach, she leverages her vast tech support background to provide accessible and insightful coverage on latest technology trends.
Les mer »
Google’s Gemini AI now has a new app and works across Google productsGoogle is getting rid of Bard and Duet, and it’s giving all its generative AI products a new name: Gemini. It’s also launching Gemini Ultra in a new $20 a month subscription.
Les mer »
Gemini Advanced Is a Central Part of Google’s Subscription FutureGoogle is still mostly supported by ads, but its new subscription-only chatbot, Gemini Advanced, is a major addition to the company’s burgeoning services business.
Les mer »
Google Has Renamed Bard After Gemini, The AI Model That Powers ItI'm a Forbes technology reporter and senior writer based in San Francisco. Send me tips at rnievaforbes.com or richardnievaprotonmail.com.
Les mer »
How to Get Gemini Advanced, Google's Subscription-Only AI ChatbotGoogle's answer to ChatGPT Plus gives you access to Gemini Advanced, its most powerful chatbot yet. Here's how to subscribe to get access right away.
Les mer »
Google’s Gemini assistant is fantastic and frustratingGoogle’s Bard chatbot is now Gemini, and you can set it as your default voice assistant. It shows a lot of promise in getting things done, but it’s not the best conversationalist.
Les mer »