• AInauten.net
  • Posts
  • 🦙 The new Meta AI for 3.6 billion people ...

🦙 The new Meta AI for 3.6 billion people ...

👉 PLUS: New deepfakes and robots

This issue is brought to you by:

AI-HOI AInauts,

Welcome to the latest issue of your favorite newsletter.

That's what we have in store for you today:

  • 🦙 Attention, Llama 3 spotted in the wild-
    on Meta.ai, FB, IG, WhatsApp & Open Source for all!

  • 🪢 AI quick tip: Tidy up your browser tabs

  • 😲 Microsoft VASA: Deepfakes more real than ever before

  • 🦾 Atlas, the humanoid superhero robot

  • 😁 Fun Times - but only if you're an AI ...

Let's go!

🦙 Zucks Meta Move: Llama 3 new on Meta.ai, Facebook, Insta, WhatsApp & Open Source!

Boom! We knew something was coming … but we weren't prepared for how hard it would hit!

MMA fighter and Meta CEO Zuck has not only presented a new gold chain, but also a latest-generation llama. Check out all details here.

The most important facts in a nutshell

Meta is integrating its new AI model Llama 3 into a standalone app at https://meta.ai and into the entire meta-universe, from WhatsApp to Instagram and Facebook - with Ray-Ban glasses and Quest headsets soon to follow. (No access yet? Use a VPN with US IP).

In plain language: The rollout will be gradual, but the new Meta Chatbot will very soon support the interaction of 3.6 billion (!) users worldwide.

For many people, this will be their first encounter with the world of AI. And Meta makes no compromises here, because the bot is being pushed for all users.

Image and animation via Meta.ai

An open source model at GPT-4 level (almost)

Meta not only incorporates this into their own portfolio, but also makes the model freely available to the public. You can even install it locally.

Really cool - and actually, this is what OpenAI originally promised! Developers can download it here.

Llama 3 comes in three versions: 8B, 70B and 400B (still in training) and promptly asserts itself as the alpha among the open source models.

In the leaderboard of the “Chatbot Arena”, the 70B version has already pushed ahead of Anthropic's Claude and Google's Gemini (at least in the “English” category).

Did you know? You can let all chatbots compete against each other in the arena - and use them 100% free of charge!

The chatbot in a practical test - promising, but also pushes the limit …

The chatbot masters all common tasks (texting, programming, creating images, etc.) with ease.

A few features stand out:

  • Meta AI has access to the web, and you can get up-to-date information in real time, via Google and Bing connection! That's cool.

  • The output and interaction in languages other than English is still a bit bumpy.

  • Facebook, Instagram, or WhatsApp - Meta AI is with you along the way. You can search for information directly in your apps without having to leave them.

  • Not only can you create images, you can also customize and animate them superfast.

  • And if you start your chat on meta.ai with “Imagine”, the images will be generated continuously while you are still typing.

  • Note: All images have a watermark (bottom left) and can currently be generated in square resolution, but not yet in very high resolution.

  • Meta emphasizes security and responsible use. So far, so good ... but we were a little surprised when we were able to generate Bart Simpson, SpongeBob or “the president in a compromising situation”.

  • But that's not all: with the right jailbreak, illegal instructions for all sorts of things can also be extracted from meta.ai. Mr. Breaking Bad Walter White would have had some fun with this …

Our take: The most important milestone since ChatGPT!

With Llama 3, Meta is definitely setting a new standard for the integration and accessibility of AI. Therefore, this step marks a significant milestone not only for Meta, but for the entire humanity industry.

Zuck is democratizing AI technology and giving billions of users around the world access, regardless of their technical expertise.

While there are legitimate reservations, we find Meta's commitment very promising. And this is just the beginning …

And between you and me: OpenAI will certainly follow up very soon with GPT 4.5/5 to get some of the spotlight again. This will be really exciting! … until then you can listen to this interview.

🪢 AI quick tip: Organize your browser tabs

Question: What is this?

Answer: Exactly, our browser with 117 open tabs when we write a newsletter ...

There is a simple solution for more order and clarity - with two clicks you can group the tabs into suitable topics, thanks to AI:

The result: Thematically grouped tabs!

Spoiler: If the tabs are all on the same topic, you may still end up with 50 tabs in the same tab after organizing them …

But here is the prime example.

Next Step: Just try it out!

  • You can find instructions for Microsoft Edge here (Edge is the best AI-powered browser anyway, in our humble opinion - give it a try!),

  • details for Google Chrome here (not yet available in the EU, VPN?),

  • and instructions for other browsers here.

😲 Microsoft VASA: Deepfakes as real as never before

Microsoft has teased an exciting technology with VASA-1.

Recipe: Take a single portrait photo and an audio clip and mix them together. Stir three times and the deepfake video is ready!

Don't you believe it? We've copied everything together, here's the video - see for yourself! The rapping Mona Lisa has a lot going for it …

These videos are not only impressively in sync with the audio, but also show lifelike facial expressions and natural head movements that can be created virtually in real time. Wtf?

The innovation behind this is the ability to generate subtle facial nuances and head movements that are deeply synchronized with the spoken word.

Our take: We are already in the middle of the era of deepfakes ...

However, the risk of misuse by deepfakes is very serious here ... Microsoft has made it clear that they want to prevent that, and are therefore not currently planning a public release.

But it's only a matter of time ... because there are plenty of alternatives that aren't quite as good, but almost.

🦾 Atlas, the humanoid superhero robot

If these deceptively real deepfakes have got you thinking, then wait till you see the new Atlas humanoid …

After backflips and dance performances have amazed the internet in the past, Boston Dynamics has unveiled a new version of the Atlas robot.

In the video, he shows off his impressive skills. Thanks to flexible knee joints and a torso that can rotate 360 degrees, he moves in a way that no human can imitate.

With this agility, even Cirque du Soleil athletes will be envious …

The new Atlas not only stands out from its predecessor thanks to its mobility and its electric drive. It can also move heavier loads than most people and the previous hydraulic version.

While the original Atlas was intended for search and rescue missions, the new Atlas is also designed to help transport bulky and heavy objects quickly through warehouses.

It will initially be tested in the factories of Boston Dynamics' parent company Hyundai before being sold commercially.

Our take: Atlas shows where the robo-journey could take us

Atlas is not just a technical demonstration of power; it is also a glimpse into the future of robotics. With such advancements, Boston Dynamics is pushing the boundaries of robotics technology into new realms.

And it clearly shows that the best robots don't just have to be modeled on humans to cause a stir in the world of humanoid robotics!

😁 Fun Times - but only if you are an AI …

No comment needed.

We made it! But no need to be sad. The AInauts will be back soon, with new content for you.

Reto & Fabian from the AInauts

Follow us on Twitter & LinkedIn!

Your feedback is essential for us. We read EVERY comment and feedback, just respond to this email. Tell us what was (not) good and what is interesting for YOU.

🌠 Please rate this issue:

Your feedback is our rocket fuel - to the moon and beyond!

Login or Subscribe to participate in polls.