Google Cloud today announced a flurry of AI announcements, including new models available in Vertex AI, upgrades to the Gemini API, and new languages supported by AI in Google Translate. Developers can now take advantage of the 2 million token context windows in Gemini 1.5 Pro without having to wait on a waiting list. Plus, you can now apply to be one of the limited number of users of Imagen 3, Google’s latest image generator that lets you create photorealistic images for marketing or corporate presentations.
Google’s latest AI is now open for business.
Today, several new or higher-performance models from Google AI are more widely available on the Vertex AI platform.
- Gemini 1.5 Flash, a relatively compact model with a million token context window, is generally available.
- Gemini 1.5 Pro is generally available.
- Imagen 3 is in preview. Apply here.
“Gemini 1.5 Flash makes it easier to continue scaling up to apply generative AI to high-volume workloads without compromising the quality of the output or context window, even for multimodal use cases,” JC Escalante, global lead for generative AI at market research firm Ipsos, said in a Google press release.
Vertex AI currently offers or will soon offer:
- Lightweight Gemini variant Gemma 2: It will generally be available in two sizes, 9 billion parameters and 27 billion parameters, from Vertex AI next month.
- Anthropic’s Claude 3.5 Sonnet, now available for purchase.
- Context caching, a technique used to generate higher speed and lower cost for AI requests using repeated content, is now in public preview for Gemini 1.5 Pro and Flash.
- Provisioned throughput, a feature of Vertex AI for provisioned workloads on the Gemini model, is now generally available to whitelisted users.
- Grounding for better accuracy is now available, allowing AI to compare information to Google searches. Groundings from third parties such as Thomson Reuters will be available starting next quarter.
- Grounding using High Fidelity mode, which combines Gemini 1.5 Flash with corporate data, is currently in experimental preview.
Vertex AI is available in a variety of geographic regions.
Note: Here are five ways to use generative AI to search the web.
The Gemini API can now run code and perform other tasks.
Code execution is now possible on Gemini 1.5 Pro and Gemini 1.5 Flash, allowing developers to run Python inside their models and experiment with generative AI iterating and learning from the code. It can be accessed through the Gemini API or Google AI Studio.
Additionally, Gemini API users can now:
- Get the full 2 million token window in Gemini 1.5 Pro.
- Enables context caching on both Gemini 1.5 Pro and 1.5 Flash.
- Experiment with Gemma 2 in Google AI Studio.
Add Cantonese and 109 other languages to Google Translate
Google has added 110 languages to the public Google Translate service using the PaLM 2 language model, the largest expansion to date for the service. The highlight is Cantonese, a language that Google has had trouble finding data to add to Translate in the past because it “often overlaps with Chinese in written form.”
PaLM 2 allows Google to add languages that are similar to each other more efficiently, Isaac Caswell, a senior software engineer at Google, said in a press release about this expansion of Google Translate.
