🔥 Weekly AI news: Did you miss it?!

👨‍🚀 The most important AI updates at a glance

AI-HOI, AInauts!

Maybe you didn't catch all the news, tools, and hacks about AI last week, or maybe you've only recently joined us. Either way, here's our recap with all the headlines from the newsletter - just one click away!

Click the links to jump right to the article - or read our picks below.

• 🤴 Frighteningly good: what the Claude chatbot can do for you

• 👾 New free tools for AI avatar videos

• 🇪🇺 EU's AI blunder - How Brussels screwed up our tech future

• 👮 Is the ultimate surveillance coming with AI?

• 📹 AI avatars show emotions

→ Selection of the top posts of the last week ←

🤴 Frighteningly good: what the latest Claude chatbot can do for you

If we were the New York Post, we would start something like this: This new artificial intelligence is incredibly good! It will replace sooo many jobs (including yours!), and very few people have any idea what's in store for them...

But we're not the New York Post. Enough clickbait and hype, because don't worry: it's not (quite) like that!

What is it all about?

Until a few days ago, we thought ChatGPT was the best of all worlds. But as soon as we got our hands on the new Claude Sonnet 3.5 from Anthropic, we gave OpenAI the cold shoulder and got excited about the new possibilities. It beats the competition in practically every respect.

But ... the best thing is the "artifacts" and the new user experience! Watch these videos.

• So, the new "Artifacts" function makes it possible to interact with and edit the results directly in the app.

• Content such as code, content or website designs are created and displayed in a separate window directly next to the conversation.

• This creates a dynamic workspace where you can see, edit and build on the results in real time. No further downloading or executing of code is required.

• In other words: Claude literally codes entire web apps for you... in just a few seconds! You have an entire (development) team in your pocket at all times.

The most important facts, and a series of examples:

• Claude 3.5 Sonnet is already available for free on the web at Claude.ai and on the Claude iOS app, with higher usage limits for Pro and Team subscribers.

• For developers, it is also available via the Anthropic API, Amazon Bedrock and Google Clouds Vertex AI (even with a Replit integration!).

• According to benchmarks, it is the "best" model in almost all areas. GPT-4o, Gemini 1.5 Pro and Metas Llama 3 400B are beaten in seven out of nine overall benchmarks and in four out of five vision benchmarks.

• Even slightly less scientific tests show the same picture.

• This includes coding, college-level reasoning, humor, image recognition, writing, translating, more human-like communication, ...

• Claude 3.5 Haiku (the little brother) and Claude 3.5 Opus (the big sister) are coming later this year, and work is also underway on features such as "Memory" to store user-specific details.

• Claude 3.5 Sonnet is already twice as fast, better and massively cheaper than its predecessor Claude 3 Opus (via API).

But benchmarks aren't everything - what counts is how the model performs in practical use.

It's super easy to access:

1. Register/create an account at Claude.ai

2. Activate the new artifact feature under "Feature Preview"

3. Insert one of the following example use cases in the chatbot - and be amazed!

Here are some possible use cases:

Claude can create documents, code, mermaid diagrams, vector graphics, visualizations, games, apps and much more.

• Upload a screenshot and create a website (from scratch)

• Web application (ChatGPT, Interview, ElevenLabs Sound-Generator, ...) or generate games (for example: Memory, Black Jack, TicTacToe, Space Shooter, Doom, Tetris, Guitar Hero, ...).

• Program a browser extension or script

• Visualize data, animate facts, explain or start simulation, create presentations, …

Here are a few prompts for inspiration:

Now it's about putting some prompts into Claude to get a feeling what's possible. As Ilya wouls say: "Can you feel the AGI?!"

Develop the most exciting blackjack game on the Internet. The presentation should be classy and professional (appealing design and color scheme, intuitive interaction elements). 

Think step by step. Make the prompt 5x better and extend the game with cool features (IMPORTANT: pay attention to your token limit! Give me the result in one-shot - no feedback, code comments or explanations, just do it!).

You are a time management professional with a PhD in behavioral psychology.

Think step by step and expand the following idea so that the result is 10x better and the user can work through their tasks as efficiently and effectively as possible! 

Create a visually appealing, interactive to-do list with a Pomodoro timer. Don't give me feedback, questions or explanations, but create it directly (one-shot - note your token limit!).

I want to find the right partner for life. You can help me if you coach me. You're a combination of Neil Strauss, the author of "The Game", and dating guru David de Angelo.

Create a visually appealing, interactive and entertaining game where you guide me through different fictional situations (flirting, dating, etc.) and give me a starting point, and then give me different multiple choice answers to choose from. After each question, give me the answer as well as feedback and tips.

IMPORTANT: Make the game 5x better and enhance it with relevant and important features that will help me find my partner for life. Always address me as you, no explanations, let's play!

Create a highly interactive, visually stunning React component of a solar system at scale. Implement this using only CSS, your available libraries, optimized for performance and educational content. Create the component in one-shot, without additional explanations or comments. 

Twitter X is full of many more exciting applications. Our friend Pliny immediately teased out the system prompt, Pietro has an agent worked out and others are already saying that Claude can make decisions just as well as the supreme court in the US (in Brazil, GPT is already making its move).

Our Take: Holy s… mokes!

A small drawback is that Claude has no internet access and can't generate/display any pictures. But we can live with that!

In 15 minutes, we "programmed" a cool browser extension that sends data to Zapier via a Webhook. Or a Google Sheets app script with which we can use with Claude, GPT and Gemini to directly add AI-formulas in the tables. ... and all done without ZERO programming knowledge!

Anthropic, which was co-founded by a number of ex-OpenAI employees, has set the bar high! The fact that Instagram co-founder Mike Krieger is now also on board as Chief Product Officer is also a big plus.

OpenAI has a lot going on at the moment ... GPT-4o doesn't quite live up to the hype (where's the new voice feature?), and there was a huge outcry when the retired US spy chief was welcomed to the board.

And the fact that former chief researcher Ilya Sutskever is founding his own company with a focus on "secure superintelligence" is not exactly conducive to trust either.

But as always in the arena of chatbots and voice models, the next surprise is around the corner - and will arrive probably sooner than we expect ...

👾 New free tools for AI avatar videos

Let's continue with the topic of videos!

After KLING, LumaLabs Dream Machine and last but not least Runway, a new tool is now in the spotlight.

We are talking about Hedra.com. You can use it to animate images with amazingly good voice output, realistic movements and lip synchronization. And it's currently free to use.

The user interface is super simple: enter your script, select the voice, upload an image (or have one generated) and poof ... in a few minutes you have your clip. Well done!

And that's not the only video news. HeyGen just got $60 million in funding, and TikTok has unveiled Symphony - a new AI suite for creators. This new TikTok tool suite helps you create content, whispers the hottest trends in your ear and turns your thoughts into cool scripts.

The Symphony Creative Studio also dreams up videos that look like you've been working on them for days. You can even use AI avatars that look like real people and talk in different languages.

And for the advertisers among us: Symphony also has a few tricks up its sleeve to spice up your ads. Auto-diagnostics, optimization, new variants - no problem! It all sounds pretty exciting, and we're waiting for it to go live.

🇪🇺 EU's AI blunder - How Brussels messed up Europes future

We have to say it again: The EU has really screwed up!

Apple just announced that the new Apple Intelligence features are not coming to Europe, and Meta has also halted the rollout of its AI tools.

And this is not the first time! In Europe, we're still waiting for the latest features from ChatGPT, which have long been available in other countries. At least they haven't banned VPNs yet.

While the rest of the world is taking off with AI, Europe is missing the boat!

And why? Because the EU's Digital Markets Act (DMA) is causing headaches for the tech giants. The DMA was supposed to be the secret weapon against the superiority of Big Tech.

Instead, it has mutated into a regulatory boomerang. It is worded so vaguely that tech companies are scratching their heads. And it forces them to divulge the inner workings of their AIs and share them with the competition.

In was an attempt to protect our data and ensure fair competition. But instead, the EU has inadvertently blocked our access to the future. It's like banning bicycles to prevent traffic accidents - sure, problem solved, but at what cost?

What is left for the poor, innovation-hungry Europeans?

The EU urgently needs to rethink and revise the DMA. It is time for the EU to foster an environment of innovation while protecting the rights of its citizens.

Who knows, maybe today's glitch will become tomorrow's feature. Until then, keep your fingers crossed and hope that Brussels finally wakes up from its digital slumber!

👮 Is the ultimate surveillance coming with AI?

We usually try to have a neutral and objective lens, and the post above was a clear statement. Sometimes, you have to take a stand 😎.

But the next topic shows, why it is perhaps a good thing that not all AI trends and tech gadgets are approved and available right away.

When it comes to AI risk, there have been a few things that can make you feel a little queasy.

Topic: Surveillance

Just recently, OpenAI announced that the former head of the US National Security Agency NSA takes a seat on the board. Many were not happy about that.

And now, OpenAI just announced that they’ve bought a new company (no, not database analyst Rockset - that was acquired last week) called Multi ...

Multi offers the possibility to control computers remotely!

Sure, it fits perfectly into the product strategy. And somehow we want to have the feature after we got a taste for it with the Mac app (finally, it is available-to-all). We can talk to the AI while it operates our computer and does things directly for us - like a smart assistant.

Nevertheless, in this context, it makes you think about the implications.

The last item on the list is a presentation is already a few months old, but illustrates very well what we can expect with AI.

"AI turns Wi-Fi routers into cameras."

The presentation will show how an AI model was developed with the help of real cameras and WLAN routers, which can now be used to represent people in a room without cameras and only using radio frequency signals.

What does that mean?

WLAN routers are everywhere. Wi-Fi routers are also constantly being hacked. And with such Wi-Fi routers and AI, it is possible to detect whether people are in the rooms, at which location, etc.

The AI can see through walls without a camera ...

Kind of spooky … 👻

The whole presentation on AI security is quite interesting. The creators behind it, around Tristan Harris, have already produced a lot of valuable food for thought with the Netflix documentary Social Dilemma.

You can see this in detail here:

All in all, this is not meant to be a doomsday forecast. But we don't just want to highlight the positive developments, we also need to talk about the risks ans potential downsides.

Because if you add up the many developments, there is definitely an argument to be made that the AI carries risks if not handled ‘properly’ (whatever that may be).

Your AInauts, Fabian & Reto

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.