Google Strikes Back 2024: Astra vs. GPT-4o, Gemini Revamps Search
Google Goes Wild: Releases Nearly 10 Models Overnight!
This May feels like a throwback to March 2023, with a series of exciting AI events unfolding.
Last March, Google released the PaLM API. At the same time, OpenAI launched GPT-4 and grabbed all the attention. A few days later, Microsoft announced GPT-4 integration into Office, overshadowing Google’s efforts.
A similar situation is happening this year. OpenAI unveiled the upgraded GPT-4o yesterday, and Microsoft is set to host Build 2024 next week. Can Google overcome this competition? We’ll get a glimpse at the I/O 2024 developer conference, which started early this morning.
This I/O conference marks the eighth year of Google’s “AI First” strategy.
Highlights
Google CEO Sundar Pichai recently stated in an interview that “AI is still in its early stages. Google will eventually win this war, just like it did with search.”
At the event, Sundar Pichai shared the latest updates from Google:
The Gemini app is now available on Android and iOS. With Gemini Advanced, users can access Google’s most powerful model.
Google is launching an improved version of Gemini 1.5 Pro to developers worldwide. Gemini 1.5 Pro, with a 1 million token context, is now available to consumers through Gemini Advanced. It supports 35 languages.
Google has expanded Gemini 1.5 Pro’s context window to 2 million tokens. It is available to developers in a private preview.
Google is looking into Project Astra. It’s early for Agent, but Astra uses phones to see, read, find glasses, and hear sounds.
Gemini 1.5 Flash, a lighter version of Gemini 1.5 Pro, has been released. It’s optimized for low latency and cost-sensitive tasks.
The Veo model for creating high-quality 1080p videos and the text-to-image model Imagen 3 have been released.
Gemma 2.0, featuring a new architecture and 27 billion parameters, has been introduced.
Android has become the first mobile OS with a built-in foundational model, deeply integrated with Gemini, making it an AI-centric OS by Google.
The sixth-generation TPU, Trillium, has been released. It offers 4.7 times the computing power per chip compared to the previous TPU v5e.