THE NEWS

🤖Google unveils ChatGPT competitor Gemini

Let’s start with the good stuff:

Gemini is a multi-modal AI that is capable of awesome things if we can believe their hands-on demo video. It understands your doodles, creates games from whatever you show it, solves puzzles, and answers tough questions. It's like having a real-time chat with a fellow human – seriously, it’s that good. You should totally check out the video:

GPT-4 has upped its game a couple months ago with adding models for vision and audio capabilities, but Gemini is built different. It’s designed from the ground up to handle multiple modes of input, which Google says makes it smoother in switching between different tasks.

It’s pretty neat, actually. Gemini can read and spot errors in your handwritten physics homework, walk you through the solution, and even suggest extra practice problems.

And there’s more: Gemini can whip up a UI to help you brainstorm and organize complex ideas. Imagine planning a birthday party and having all your ideas laid out neatly instead of a jumbled wall of text. It’s a game-changer for how we interact with AIs.

Gemini comes in three sizes: Ultra, Pro, and Nano. Nano is for mobile devices, Pro is a bit better than GPT-3.5, and Ultra is the top-tier model that slightly outdoes GPT-4 in most benchmarks.

BUT: Google hasn’t released Gemini Ultra yet, and it won’t be out until next year.

Google launched their most anticipated AI and showcased all these cool Ultra features, without actually launching the Ultra itself. Yep.

The reason: It’s not safe yet.

But at least they have rolled out Gemini Pro (which is pretty much like GPT3.5) to Bard in its full capacity. Right?

Well…except:

• It only works in English.

• It’s not available in Europe.

• And it’s not multi-modal.

And guess what? None of the Gemini versions can generate images at the moment. Google plans to add this feature later on. It seems their model should be capable of generating images natively instead of using a separate tool like ChatGPT does with DALL-E 3. But that’s probably trickier to do while keeping things safe.

On the bright side, Gemini is free, just like GPT-3.5 and Claude have been. But if you want the best free AI right now, Bing’s GPT-4 powered Creative or Precise mode are where it’s at.

The cool part about the updated Bard is that it’s hooked up with other Google stuff, so you can search your emails and calendar with the help of the AI. If you have a Pixel 8 Pro, you’ll soon be able to use Gemini Nano for things like summarizing in the Recorder app and smart replies in WhatsApp.

In conclusion: Google's release is a classic example of today's tech scene – tech giants launching products with key features 'coming soon’. It’s underwhelming that Google who is supposed to be the main competitor of OpenAI could only release a GPT-3.5 level AI in 2023.

The demos of the Ultra are promising, and would be game-changing, but it’s unclear what Google means by ‘early next year’, and where will OpenAI’s technology be by that time.

ONE MORE THING

Gemini Pro vs GPT-4 in a code challenge

Here’s a first comparison between Bard and ChatGPT in a UI design code challenge.

⌚ If you have one more minute

• 🍎 Apple launches MLX machine-learning framework for Apple Silicon.

• 📉 C3.ai’s stock drops as AI software company misses on top line

• 💬 Jailbroken AI Chatbots Can Jailbreak Other Chatbots

AI Art of the Day 🎨

Chandelier Video clip turned into liquid with AI by redditor u/qstone75. Watch the whole thing here.

🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄

That's all for today, folks!

If you enjoyed this, please consider sharing this hand-crafted newsletter with a friend.