AI-HOY AInauts,

Happy holidays and welcome to the lastest issue of your favorite newsletter.

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

• 🤯 Breaking news: OpenAI unveils o3 - is it AGI yet?

• 📺 AI video update: Google Veo 2 is really this good ...

• 👓 ChatGPT + Ray-Ban Meta glasses = your perfect guide for on the go

• 🖊️ Fun: The ultimate selling point

Let's go!

🤯 Breaking News: OpenAI unveils o3 - is it AGI yet?

OpenAI has ended the 12 days of Christmas surprises with a real highlight... Imagine waking up one morning and your AI assistant is suddenly as smart smarter than an entire team of Harvard professors! That's exactly what OpenAI promises with its latest model, o3.

But first things first: the model is not yet publicly available, but safety testers can apply here. According to Sam Altman, the launch of o3-mini is planned for the end of January and the full o3 model shortly afterward.

But is it really the long-awaited and dreaded breakthrough to AGI, artificial general intelligence? Let's see...

P.S. If you're wondering why o3 comes directly after o1 and why there is no o2: there's a telco company that has already successfully established itself under this name … 😁

The cool feature of o3: it's incredibly smart!

The latest models don't necessarily focus on even better pictures, videos or audio, but above all on the ability to think about really difficult things and find solutions. The same applies to Google, which has just introduced a new "thinking model".

The sticking point with the latest models is therefore that there are fewer and fewer tests that can be used to objectively check how smart the model really is.

o3 achieved a result of 87.5% in the infamous ARC-AGI test. That is better than comparable human scores. For comparison: GPT-4 had a meager 5% at the beginning of 2024. That's like your old Polo suddenly leaving a Ferrari behind on the highway! A lot of computing power worth six- to seven-figures was made available for this purpose…

The system can even flexibly adjust its "thinking time". For simple tasks, it provides answers every millisecond; for complex problems, it takes its time to ponder.

Under the hood is a new NPO architecture ("Neuronal Pathway Optimization"), which is based on the human brain. This means that o3 no longer processes information like a mindless data store, but actively connects and combines knowledge - almost like we humans do.

Next Level Skills: What o3 is really capable of

o3 beats many benchmarks by a lot. To illustrate this a little more vividly:

• Coding: o3 programs better than 99.9% of all humans and lands in 175th place among the world's best programmers! In the SWE-Bench Verified Test, o3 scored 71.7%, which is 22.8 points higher than its predecessor o1. Absolutely awesome...

• Science: It spits out extremely high-quality scientific papers. o3 scored 87.7% on GPQA Diamond, well above the typical level of PhDs, which is around 70%.

• Math: o3 solves math problems better and faster than math professors. It achieved an almost perfect score of 96.7% at the American Mathematical Olympiad (AIME). On EpochAI's Frontier Math, o3 solved 25.2% of the problems, while previous models didn't even crack 2%.

o3 can combine knowledge from different areas and generate new solutions. This is no longer just blunt pattern matching, but real, creative problem-solving that can be applied to all possible areas and disciplines.

What we find particularly striking is the massive improvement from o1 to o3 - and how the average person performs in these tests …

Our take: Science fiction becomes reality

o3 is cool tech with two faces! On one hand, it pushes scientific breakthroughs and innovation forward in turbo mode. On the other hand, it brings tough challenges in terms of jobs, skills and control.

It is definitely the next big milestone in the evolution of AI. But is it a "real" AGI yet? The question is almost philosophical, because for most of us neither o1 nor o3 or o4 are going to make a difference in our daily business.

But that could change as soon as the handcuffs are taken off the model, and it can take over real tasks for us and control the computer or autonomous agents. This will definitely usher in a new era of human-machine collaboration!

Perhaps this will happen as early as 2025. If developments continue at this rate, humanity will soon not only have AGI, but even ASI, Artificial Super Intelligence.

What that means is beyond human imagination (at least ours). We therefore recommend that you take a look at Ray Kurzweil's predictions and internalize the articles by Sam and Dario from Anthropic.

And then it's time to buckle up and head full throttle into 2025! We are still at your side as co-pilots and are looking forward to it.

📺 AI video update: Google Veo 2 is really this good ...

Kling, Sora, Runway, Dream Machine, Minimax etc. The variety of video models is plenty, and they are making progress from week to week. Google's new Veo 2 model, however, is currently making a name for itself, as these gems illustrate. You can join the waiting list here if you also want to play with it (use VPN with US IP, if needed).

👓 ChatGPT + Ray-Ban Meta glasses = your perfect guide to go

The Ray-Ban Meta glasses are by far the best investment we've made in the last year! Leaving the house without these glasses has become an absolute no-go. Reason enough to give you a few of our practical use cases as inspiration.

What can the Ray-Ban Meta do?

It is a stylish pair of glasses that not only looks good, but is also equipped with artificial intelligence - the perfect symbiosis of style and technology.

They look like normal Ray-Bans, but are equipped with microphones, speakers and cameras that are seamlessly integrated into the design. You can have them as sunglasses or with normal lenses, even corrected ones.

Your personal AI assistant always at your side

• With the integration of AI in the Meta Ray-Ban glasses, you have a personal assistant right in front of your eyes. You can use the built-in Meta Llama assistant or talk to the ChatGPT app on your phone via Bluetooth.

• Whether you need a quick answer to a question, want real-time translations, are on the phone with someone or are looking for information about your surroundings - it's all possible.

• Thanks to the Bluetooth function and the built-in speaker, you can listen to music (with native Spotify integration) or make calls (with WhatsApp even with video!) without the need for additional headphones.

• THAT is a real highlight! The audio quality is impressive, and you stay in touch with your surroundings at the same time. Super practical, if only to have background music in your ear - and if the volume isn't too high, no one will even notice.

• You can take photos and videos with one click - or even stream live. We love taking snapshots and short videos to capture everyday situations without having to take out our phones.

• The battery life is pretty good, and if necessary, you can quickly recharge the glasses at any time using the stylish charging case with built-in battery.

• And Meta is constantly delivering new features - Live Translation, Shazam integration and new video AI functions have just been introduced.

Our take: Treat yourself to a Ray-Ban Meta - best Xmas gift ever!

The Meta Ray-Ban glasses with ChatGPT integration are more than just a gadget - they are a step towards a world where technology is seamlessly integrated into our everyday lives. It gives you the freedom to enjoy information and entertainment without being distracted or having to stare at screens.

We often use the ChatGPT assistant. For example, last time in the museum we started an audio conversation and asked ChatGPT Voice to tell an entertaining story about the artist and what his surroundings looked like at the time. And in the meantime, ChatGPT Advanced Voice Mode can even see and provide you with even better information.

This allows you to explore your topics and interests in a very individual and exciting way. It's like your own personal guide. Of course, this can also be applied to other environments, such as a city or any topic that interests you.

WE LOVE OUR META RAY-BANS - thanks Zuck! ❤️

🖊️ AI-Fun: The ultimate selling point ...

The simplest sales argument of all time: "It's AI-powered!"

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

Reto & Fabian from the AInauts

P.S.: Follow us on social media - that motivates us to keep going 😁!
Twitter, LinkedIn, Facebook, Insta, YouTube, TikTok

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.