Google I/O AI keynote updates 2024

This is a rundown of some of the main features of the event.

Hello Everyone,

I was more interested in Google I/O this year than usual in 2024. Google like OpenAI, really hit home with multi-modality and longer context windows. But it’s the multiplicity of Google’s product upgrades that is so multi-faceted, fascinating and frankly confusing.

Google is putting AI into everything in their existing products, but also releasing many new AI related products. I’ll be presenting a series of articles about this, since there’s a lot to unpack.

---

📶 From our sponsor: 📶

The AI Agent playbook for founders and operators building agents

If LLMs are to extend beyond mere Q&A, they need:

1. A suite of Tools to actually impact the real world

2. Memory repositories to remember what actions they’ve taken

3. Auto-critique algorithms to error correct along the way.

Then comes an agent.

A Technical Guide for Founders & Operators Building Agents - download the report.

---

Can Google Search Reinvent itself with SGE?

Google Search Generative Experience is a set of search and interface capabilities that integrates generative AI-powered results into Google search engine query responses.

• This summary of Google’s developer conference will include a lot of images and short videos so you can get a quick idea of what it was about and dig in deeper if needed.

• More than ever, Google I/O is directly related to Google’s product updates that have to do with A.I.

Support and Dive Deeper

Support my work for as little as $2 a week. Get access to deep dives and AI related report summaries. 🎓📚💡

Highlights

Hint: Opt-in to waitlist at labs.google. to get some early access to some of its newer features.

• AI Overviews

• Ask Photos: Can make voice queries with Google Gemini of your Google Photos

• Gemini 1.5 Pro’s huge context window (e.g. 1 million context window)

• Context Window is expanding to 2 million tokens (making available to developers in private preview)

• Google Workspace with Gemini

• Notebook LM with Gemini 1.5 Pro now with Audio Overviews

• Gemini 1.5 Flash

• Project Astra: a Universal AI Agent, see the demo (YouTube).

• Imagen 3: Image Generation (Sign up in Image FX)

• Generative Music: Music AI Sandbox

• Video Generation: Veo (new experimental sandbox called Video FX)

• Trillium, 6th generation TPUs.

• Google Search: now 1 trillion facts, optimized with a Models specifically for this.

• Asking Questions with Videos directly in Search: Opt-in to Labs to be among the first to see new features in Google Search

• Upgrades to how Gemini can interact with Gmail and data there.

• Gemini AI Teammate

• Gemini App, and upgrades

• Gemini Trip Planning

• Gemini Advanced 1.5 Pro with 1 Million Context windows (e.g. you could upload a PDF with 1,500 pages)

• On device AI for example to protect from Scam phone calls

Google for Developers

http://ai.google.dev/

Build with state-of-the-art generative models and tools to make AI helpful for everyone.

• Build with the Gemini API

• Also see, AI Solutions for Developers

Just a Note: What is Gemini Advanced?

You can upgrade to Gemini Advanced, which is part of a paid plan in Google One that also includes:

• Gemini in Gmail, Docs, and more

• 2TB of storage

• and other benefits

Upgrade to Gemini Advanced

To upgrade to Gemini Advanced, subscribe to the plan in Google One.

1. Go to gemini.google.com.

2. At the top, tap Menu

   

3. At the bottom, tap Upgrade to Gemini Advanced.

4. Follow the on-screen instructions.

5. It is $20 a month.

🎧 Audio Intro: 2:04

Introduction

With the AI craze in high gear, expectations were high for Google this year. Some of the most incredible papers still come out of Google DeepMind. Like RecurrentGemma. Google DeepMind published a research paper that proposes language model called RecurrentGemma that can match or exceed the performance of transformer-based models while being more memory efficient, offering the promise of large language model performance on resource limited environments.

Google I/O is full of applied and half-baked products Google is working on, and you could make the argument Google releases too many models, products and intangibles to fully grasp all that they are doing. Google still feels like an AI playground of sorts and that’s both the charm and weakness of Alphabet during this Generative AI era of hype and sky high expectations.

Google I/O keynote day still thrills, well because, it’s Google. With all that Microsoft and OpenAI have done, you’d hope that Google can bring us closer, tangibly to an AI unified future and not just boast about its Multimodal AI or its multiplicity of models. Unfortunately, that’s not the case here.

Google kicks off its developer conference each year with a rapid-fire stream of announcements, not always connecting the dots to how consumers will actually use or apply them in real life, not to mention even the office! By going big, nothing feels very deep with Alphabet, or as substantial in an applied way as Microsoft Copilot Era has successfully undertaken. Even Google Gemini, feels, well, almost transcendent.

And it’s not that I have any problem with AI ad nauseam products, I just want to sense that they are actually useful to me and the real world. To be immersive AI products need to have reduced friction and increased accessibility, and I’m not sure Google is quite at even OpenAI’s level with this element of product-market fit. But that being said, let’s dig in shall we?

Google I/O is Google’s annual developers’ conference. It often gives sneak peaks on new consumer products that are coming as well.

💡 Google I/O 2024

What the 2024 Keynote

If you want to watch the entire thing, here you go: (the event was 3 hours long)

🔎 OK GOOGLE

What will a 2 million context window bring to the world?