Google's AI plan is a 4 ingredient recipe 🍔

Google had its I/O event on Wednesday and they were ready to dominate the AI news cycle.

They shared A LOT OF news, but 4 specific announcements outlined their AI strategy for the year.

They’ll focus on the scalability of language models, multi-modality, music, and AI safety.

Let’s take a look.

🌴 The launch of PaLM 2

PaLM 2 is a suite of 4 language models which excel at sense reasoning, mathematics, coding, and logic tasks. Some models are so lightweight that can run on mobile devices.

“What we found in our work is that it’s not really the sort of size of model — that the larger is not always better.”

They saw how open-source is making these models accessible for every device/use case and are trying to replicate that.

To use the API get on the waitlist.

♊️ The Gemini model

Gemini is their multi-modal model, that’s still in training. It’s also highly efficient at the tool and API integrations and built for long-term memory.

Despite being late to push a model like this to production, Google shared that their Gemini model demonstrates remarkable results even during training, surpassing current market models.

We’ll confirm this when they launch.. (ahem) 😅

🎸 MusicLM

Finally, what I am really excited about is the launch of MusicLM in their “TestKitchen” app.

This is the text-to-music model that they announced in January. You can sign up to use it and you can generate music from text for free.

✏️ Image watermarks

All AI-generated images from their models will have embedded watermarks and other techniques to identify if an image is AI-generated.

Their stance on AI is that we must be able to identify AI-generated from real images.

A quick recap of the rest:

1. Chat-style answers to Google search for US only (link)

2. Android Gets an AI Boost (link)

3. Adding AI in Docs, Sheets, and Slides (link)

4. Bard is now available to everyone (link)

   1. You can export generated text to Google Docs and Gmail.

   2. More visual (the results are text + images)

   3. Bard plugins are coming too

   4. Link to try Bard

5. “Help me write in Gmail” feature (link)

6. Magic editor for photos 👇🏻

WHAT ELSE IS GOING ON

🦙 OpenAI is working on My Profiles. They’re still working on this (not yet a live feature), but you can essentially add your personal data so that ChatGPT will forever remember that for every conversation you have.

👀 Anthropic extended their context window to 100K tokens or around 75K words. You can now upload a whole book and chat with it. Still not public, and is available only through their API - you can get on the waitlist here. I wonder what this means for vector database companies like LangChain 🤔

🏋🏻‍♀️  Stable Animation by Stability.AI. The creators of Stable Diffusion just released their model for text-to-animation. You can create animations via prompts (without images), a source image, or a source video. Here’s how to access/ use it.

💰 ChatBase broke 60K MRR. I’ve never seen growth like this since the AI Avatars season. This is an app that creates custom ChatGPT for your website. And there are dozens like it out there. Wild!

RESOURCES

The best resources we came across lately that will help you become better at writing prompts & building AI apps.

📚 Upload and search your docs with No-Code [ 30min video tutorial ]

👋🏻 All you need to know about Prompt injection [short article]

🎥 Plan and Execute Agents [ Langchain agents inspired by BabyAGI]

TOOLBOX

The latest AI tools to use or get inspiration from.

• FineTuner.ai - build your own AI agents, 0 coding needed

• RoamAround - your personal travel planner

• OpenOs - ChatGPT for your financial data

• Albus - ChatGPT for your Google Drive data

• Jobed - Job description generator

• Raply - AI-powered email marketing

• Molin - Multilingual AI copywriter for free

• 3D prompt - Create funny 3D art from text

PROMPT OF THE DAY

TOOL

Midjourney

PROMPT

50mm cinematic, his introspective style, her sensitive crooning, their golden touch --ar 3:2 --v 5.1

RESULT