🔥 Weekly AI news: Did you miss it?!

AI-HOI, dear 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.

• 📣 Next level storytelling: Script in, video clip out!

• 🗺️ The MAD map: The AI jungle at a glance

• 📰 AI news quickie: the HAI highlights

• 🤭 AI-Fun - the automated BAIby car

• 🤖 This AI can do more than just speak, hear and see

• 🤓 Why everyone should work with AI APIs

• 🎧 New favorite (AI) app

• 📰 AI news quickie: the HAI highlights

→ Selection of the top posts of the last week ←

📣 Next Level Storytelling: Script in, clip out!

Y Combinator, the talent factory of the tech scene, has struck again! This time, it has also invested in Infinity.ai. A real insider tip that is (still) flying under the radar.

The idea is simple: Hit the keys with a script and get a movie clip!

We tested it right away, and Claude ChatGPT created a script with a few celebrities who had something to say about this newsletter …

A few hints if you want to give it a try:

• You can use Discord or a simple web user interface (alpha version) at https://studio.infinity.ai - Discord walkthrough video here.

• There is a whole library of well-known personalities available (mostly American) - currently there are over 75.

• If you use names of people who are not on the list, a random placeholder avatar is used.

• You can enter the script in different languages or have it generated - we tested it in German, and it doesn't sound bad at all (just a little bit more emotion, please!).

• You also have the option to create your own avatar for free - simply upload a two-minute video!

• The maximum length of the input script is 1000 characters (use https://charactercountonline.com).

• The output can be horizontal or also vertical and with subtitles (currently only via Discord - web is only horizontal and without subtitles).

• But beware, for the subtitles it seems to use speech recognition and not the provided script - which can lead to errors in the subtitles …

This is what the web interface looks like.

Our take: An entertaining gimmick, we'll keep an eye on it

The founders are convinced that a mini-team - without actors, directors, or other crew members - will win an Oscar within the next 10 years - and want to develop the tools to do so.

Well, there's still a long way to go - but if we look back at what was feasible a year ago and what is possible today, then it seems… possible?!

In a world where influencers are the new stars and TikTok videos are the only source of news, this doesn't seem too far-fetched …

At the moment, we still consider the platform to be a proof of concept and a gimmick - but it's certainly entertaining, especially when you add your own avatar to the clips and parlay with celebrities!

P.S.: Anyone interested in AI in the music & film industry should save the date for the Digital Hollywood AI Summer Summit 2024 (free registration from June 1) - and until then, watch the videos from the last summit.

🗺️ The MAD-Map: The AI jungle at a glance

Tattaaa! The new map of Machine Learning, AI & Data providers (*=MAD) has been published (interactive map here, PDF here).

It's hard to believe, but what began over 10 years ago with 139 companies on the map has grown into a dazzling ecosystem with over 2,000 players!

It's easy to lose track. That's why we look at all 2011 providers in detail, from A to Z. No worries, we’re just kidding …

But from our perspective, there are a few important insights we do want to point out.

How can you make the most of this AI-tools map?

The interactive map is super helpful if you are looking for the right tool/solution for a specific area!

In the Map view, simply enter a tool or provider in the search field at the top right, and you can see the competitors in the same box. Alternatively, switch to the Card view and find specific keywords using the full text search (Ctrl + F).

How did this diversity come about - and where will it lead to?

The insanely dense landscape is primarily the result of the massive waves of start-ups and funding rounds - and, of course, technological advances.

2023 was a testing ground for generative AI in companies. Now, in 2024, the real practical test is awaiting. We need to go beyond the showcase projects, and have tangible use cases and a measurable ROI instead - for the customers and the providers as well!

The AI critic Gary Marcus already points out, that many of the large players seem very overvalued right now, at least in comparison to other SaaS companies.

Further reading material about this topic: Will A.I. Boost Productivity? Companies Sure Hope So. in the New York Times - and if you're a data nerd or C-AI-O, here's the author of this MAD Map, Matt Turck, in an interview.

Man and machine need to work in conjunctoon - AI itself is not (yet) a solution ...

The modern AI stack is growing out of the shadow of GPT & Co. But despite all the automation euphoria, the human factor remains indispensable and the road to AI utopia is rocky.

A shortage of skilled workers, compliance concerns and integration into established IT structures require more than a Python one-liner. To score points here, you need industry expertise, implementation know how and resilience.

Unlike in previous hype cycles, the usefulness of AI applications is much higher today. Those who deliver real added value need to have no fear. But if you just slap an "AI" on top of the marketing promise, you will be discovered quickly.

Successful do not want to choose between AI and human expertise, but rather intelligently utilize the best of both worlds. This is the only way to take advantage of AI’s potential without ignoring the risks.

Our take: "AI is overhyped, but underestimated!"

The momentum is breathtaking, the progress impressive. But despite all the euphoria, we should not forget: There is still a long way to go!

That's why we think it's so important to start now and gain experience. Or to use the words of Google Deepmind boss Demis Hassabis: "Overhyped, but underestimated!"

🤓 Why everyone should work with AI APIs

We're biiiiiiig fans of everything automation.

AI in particular opens up the floodgates to an incredible amount of possibilities!

There is a specialized tool for practically every use case - but at some point you have so many tool subscriptions that it costs a fortune.

This is where APIs come into play. Almost all major AI providers offer APIs. And don't worry, these three letters are half as technical as you might think at first glance.

APIs are a type of code interface which allow you to directly access the desired models such as GPT-4, or Claude Opus etc. Providers like Replicate make many models conveniently available via API.

The big advantage of APIs in general is that you pay based on usage and are not trapped with monthly subscription fees. They are generally very inexpensive.

In addition, with a few APIs, you can quickly put together many tool functions yourself. In other words, you no longer need these pricey tools!

The big disadvantage: Until now, you had to have a technical understanding on how to work with APIs.

Now, this is where no-code apps come into play! Anyone can use them, without much technical understanding.

How? With platforms like Zapier and Make.

You can build automations and workflows - and stitch together different AI models, tools, etc. with each with a few clicks. All without programming!

Here is a current example …

A friend of ours unfortunately doesn't speak English. But he really wanted to have some English lectures translated and listen to them in German, instead of just reading them.

How can this be achieved? The easiest and best way is: ElevenLabs

Just upload the audio file, select a target language, and off you go.

The disadvantage: You need a subscription with ElevenLabs, and since some lectures are quite long, this can get pricey.

This is where APIs come in handy! We built a very simple workaround for our friend with Zapier and OpenAI, in just 5 minutes.

This is how the workflow looks like:

1. Whenever he has a new lecture, he simply uploads the file to a special Google Drive folder.

2. We use OpenAI Whisper to create a transcript of the audio file.

3. ChatGPT then translates the transcript into German.

4. After that, Whisper creates a German audio file based on the transcript.

5. This MP3 file is then uploaded back to Google Drive and our friend can listen to it.

And all we need for this workflow is a Zapier account and an OpenAI API key!

This makes a difference in terms of price. We would have paid about 5 dollars for our test with ElevenLabs. On the other, this is how much it was with OpenAI:

Thanks to the APIs, we could also easily try out Claude Opus (the best Anthropic model) instead of OpenAIs GPT, without having a monthly subscription.

This was just a quick test, but as you can see, the combination of no-code tools and APIs is an incredibly powerful one. You can automate entire companies this way!

If you are interested in this type of content, please let us know and reply to this e-mail. We want to get a better feeling if this is relevant for you.

P.S. Sure, Zapier also costs something if you want to use the so-called Premium Apps. Make is cheaper, and there are also lesser known alternatives or open source tools. Our approach is that we prefer to pay for one tool that can be used to build any type of scenario - instead of paying for 50 subscriptions, which all just cover narrow and specific use cases.

🎧 New favorite (AI) app

We debated if his app was "newsletter-worthy" …

However, since we are absolute fans and use it every day, we thought this could also be of interest to you. We are talking about Snipd.

Snipd is a podcast player that makes it extremely easy for you to digest learnings from podcasts.

At the center is a non-AI feature that allows you to save short podcast highlights that you want to remember.

In other words: if you are listening to something exciting, simply press the snip button or activate it via your headset.

Then the AI comes into play.

Snipd automatically creates a transcript of the snippet, generates a summary and gives it a title.

Other cool tool features:

• Readable and audible summaries of entire podcasts

• Chapter divisions also for podcasts that do not yet do this

• Easily export your highlights to your favorite note-taking app.

Here's the app again. (The free version is also quite usable, and available for Android and iPhone.)

We love the app, and it's a great example of how AI makes everyday life a bit better. At least for geeks like us.

See you soon with a fresh round of news, hacks and insights!

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.