🦸‍♂️ Test the best (and biggest) Google Gemini model

Good morning and AI-hoi, AInauts!

🌺 We're feeling the spring, so today is a really colorful day.

After exploring the news landscape extensively, there's a massive load of exciting updates and also a range of tools and practical tips that you can use straight away. Ready?

Here's your menu for today:

• 🦾 How to test Google's Gemini 1.5 Pro with 1 million tokens for free!

• ♟️ Personnel changes at AI pioneers cause a stir ...

• 📰 AI-News-Extended Quickie: The HAI-lights

• 😁 The Artbot is convincing - at least in sh*ttalking!

Let's go!

🦾 How to test Google's Gemini 1.5 Pro with 1 million tokens instantly for free!

Without a long suspenseful arc, let's get straight to the point: You can test Google's Gemini 1.5 Pro for free at aistudio.google.com!*

*If the Gemini 1.5 model with 1 million tokens is is not visible, try a US VPN..

There is now no waiting list, and the AI Studio user interface is a good environment for trying out and experimenting with the Gemini models (soon also via API).

The really cool thing here is the enormous context window that is available to you here. You can use it to upload entire books (series) or extensive videos - and combine your requests with a clever model that is at the forefront of many benchmarks.

Here's a quick start guide (because you can't do it without it - we'll also be making a detailed video on this in the AI Vault).

What we find particularly exciting are the use cases that are not even possible with most other models in this form (with Claude and ChatGPT, we keep coming up against upload limits because the files are too large). And if you are interested in tests for video analysis, you will find what you are looking for here.

Our take: With Google Gemini 1.5, Google has really (finally) achieved another great success! Even if not everything went smoothly during the rollout and we had to tear our mouths apart over fake videos and historically inaccurate images: all that is now forgiven.

The big advantage of Gemini 1.5 is its multimodality (text, code, image, audio, video) and the huge, massive context window. At last you can parley with someone who doesn't start hallucinating after a few minutes and forget your actual request because "the memory was full". And as I said, you can upload and analyze really large files.

We're really keen to experiment with it even more - and hope that you can also take some time to get up close and personal with "the best of the best" AI models.

♟️ Personnel changes at AI pioneers cause a stir ...

A bombshell dropped last week: the (Google) DeepMind co-founder Mustafa Suleyman has resigned from his position as CEO of Inflection AI in favor of heading a new division "Microsoft AI" (we reported).

From Google to arch-rival Microsoft is an exciting path, and many of the Inflection employees are accompanying him on this journey.

Of course, this does not mean that Microsoft is giving up on OpenAI, far from it. But Microsoft is dancing at all possible weddings and announcing deals with the most important players, from OpenAI to Mistral.

Google has also made a move and made the new Claude 3 models Haiku and Sonnet from Anthropic available on the Google Vertex AI Cloud.

But Anthropic has even more arrows in its quiver! The day after the Google update, they announced that the models can now also be used via Amazon's AWS in collaboration with Accenture. Will Google like that...?

Apropos Anthropic, the 8% stake in the bankrupt crypto scam FTX is going under the hammer - and interested parties are lining up.

All the major players are looking for strategic partnerships, as this "old" chart from January 2024 shows.

We had also expected for some time that there would be personnel changes consequences at Stability AI, after CEO Emad Mostaque repeatedly made the hit the headlines, and many key figures left the sinking ship.

Now he resigned (allegedly it was his decision), and wants to dedicate himself to open and decentralized AI in the future. What exactly this means for Stability AI remains to be seen - Stable Diffusion and co. will definitely live on in the open source universe.

Our take: The Hugginface co-founder puts it in a nutshell in the tweet mentioned at the beginning: The concentration of too much influence is the biggest AI risk of all!

And even if, as AI geeks, we are of course always excited about new features and opportunities that arise from such actions, we are aware of the risks.

The most important players are positioning themselves so strategically dominant that (open source) alternatives can hardly compete with the computing power, human talent and capital behind them ...

But let's not think about that for now, instead we'll quickly look around for the next Shiny Object - let's continue with the news section.

📰 AI-News-Extended Quickie: The HAI-lights

Finally, here are a few easily digestible news tidbits. Let's be honest: you can expect a concentrated load of hand-picked AI updates from all corners of the internet.

It's really amazing how quickly everything is developing - we are always amazed at what has happened in such a short space of time. Let's dive in!

Robotics

• The first user of Neuralink, Noland Arbaugh reveals in this video the transformation of his life through the profound impact of this technology. History!

• AI-supported robots are also on the rise! Here are the 15 most important updates from Figure.ai CEO Brett Adcock.

• Figure.ai recently presented a robot that is supported by OpenAI's models. Amazing! So it's no wonder that the who-is-who in the industry has invested in Figure.

• Tesla's Optimus still looks a bit old compared to Figure ... but Optimus and Neuralink could become a real option for people in need of care in the future.

• Or the $9/hour nurses from Hippocratic AI and NVIDIA.

• There's so much going on in the field:Covariant is building a model that robots can understand the physical world, a humanoid gropes a female presenter, Unitree's H1 does a backflip, a Chinese robot sets a new speed record and lots of other exciting updates ...

• China has developed drones that multiply in the air - and in Ukraine, autonomous drone attacks are already a reality.

• Oh yes, Meta is also busy researching and has introduced SceneScript - a model that understands 3D-spaces and can describe them.

Chatbots

• Does ChatGPT have a new personality? Until now, ChatGPT has been rather formal and concise, with a focus on information. Personality V2, on the other hand, offers a more engaging and conversational tone, and may include elements of humor or empathy. Ohh, let's see!

• ChatGPT for Teams also has new features, such as internal GPT sharing, version management for GPTs, an improved search in the store and options for DALL-E images.

• The GPT store is flooded with spam and could do with a spring clean.

• GPT-5 is set to arrive in mid-2024 after being demonstrated to some enterprise customers. It is said to have significant improvements and new functionalities. Bring it!

• OpenAI is also expanding its partner list with content publishers - in this area, the competitors still look a little old ...

• Google Gemini, Claude oder ChatGPT? Alle sind top, letztlich ist es vor allem eine persönliche Präferenz. Und eine Frage der Intelligenz!

• Eine neue Open-Source-Version von Mistral 7B, v0.2 Base, wurde am Mistral AI Hackathon präsentiert – und kann übrigens auch DOOM spielen (ok, GPT-4 kann das auch)! Kochbuch dazu gibt es hier.

  Artikel: Warum Perplexity eine innovativsten Firmen ist …

• How do ad buyers actually use GenAI? Exciting insights here!

• The Financial Times has fed a chatbot with years of material - some selected beta testers can ask it questions.

• Character.ai convinces with new voices - but oops, nice one!

• "Prompt engineering is dead!" The language models are already able to improve user input.

• Hacker researchers use ASCII art to trick chatbots. More details in the video here, and in this post there are more jailbreaks (for research purposes only, of course).

Tool tips

• Browsers are a personal preference - but Microsoft's Edge can do more than "just" AI, and we're super happy with it (even using Bing instead of Google). Give it a try and say goodbye to Google and Safari!

• Here's another geek use case: the ChatGPT predecessor GPT-2 can now be downloaded prettily packaged in an Excel spreadsheet.

• Thinkbuddy is ChatGPT for MacOS - your Mac can now do AI and listen, read and engage - so you can be more productive! We're excited to try it out.

• Freelancers who find new gigs on Upwork should take a look at the Upwork Proposal Generator "BidBuildee".

• And remember DEVIN, the autonomous AI programmer? Now there's an open source version that's trying to replicate that.

Picture & Video

• On Wednesday, March 27, Digital Hollywood will be hosting a free event on the topic of AI music - details here, registration here.

• Rightsify's Hydra is a new AI music generator that creates content for commercial use.

• Stability AI's Developer Platform offers pretty cool new editing capabilities via API! Stable Video 3D was also recently released, which allows you to generate a 3D model from a 2D image.

• In the area of video-to-video editing, there are cool research demos with AnyV2V and MOTIA Outpainting as well as VLOGGER (from Google) and EMO for AI-generated avatars - soon to be available for you too 😁 ...

• Leonardo.ai, the popular image AI, can now also generate transparent PNGs. Very nice! (For Stable Diffusion users: Instructions for transparent backgrounds here).

• You can find out how to recognize AI-generated images here.

• OpenAI is also going to Hollywood! Well, not quite yet, but there are talks with film studios and directors regarding Sora. The competition aka RunwayML, Haiper, Pika etc. is not sleeping!

• Fancy some new videos from OpenAI's Sora? Here are a few gems, and here's more.

Gadgets

• Last October at Rewind.ai we had the "most-hated-AI-device" - and are still waiting for it today... Now, with Compass is an available alternative that transcribes all conversations and runs on iOS and Android - and will also be open source!

• We love our Meta Ray Bans, and feel naked when we leave the house without them. Now, with the 32°N Muir, there is an interesting alternative for the slightly older among us - sunglasses that turn into reading glasses with a swipe.

• Microsoft introduces new AI laptops - exclusively for enterprise customers from April 9.

• 👀 👀 👀

Health

• AI helps with eye tests for diabetes patients, with breast cancer screening (better than humans!)

• Google equips Fitbit with a Gemini-based helper that can provide personalized health tips.

• Google has also come up with a AI model that can detect tuberculosis or COVID-19 based on coughing.

• South Korea has introduced AI dolls to combat loneliness among the elderly. Strange? Maybe, but if it helps you don't have to complain about it!

• The free online Healthcare 2024 Summit will take place on April 2-3.

• As a reminder: The 100 most important digital health & AI companies in 2024 from The Medical Futurist.

Government & Investments

• The US government supports Intel with 20 billion to boost chip production in the USA (and thus reduce dependence on Taiwan).

• Saudi Arabia apparently wants to provide a large AI investment pot of 40 billion and get involved everywhere.

• China, on the other hand, has invested massively in the training of AI engineers and is outpacing everyone else.

• Anguilla is also benefiting from the AI boom with the domain extension .ai - last year it accounted for 10% of the gross domestic product!

Society

• Interesting article: AI influencers are exploding - and many are also controlled by teenagers, for example via the platform 1337.org (1337 is pronounced as "leet").

• An exciting documentary by DW that sheds light on how AI will influence our future.

• Who invented it? Although not the Swiss, at least the 4th Uphill Conference will take place in Bern on May 16 and 17, 2024 under the motto "Join the AI Revolution".

• Google has predicted floods a week in advance !

• Lex vs. Sam, round 2. Listen in! ❤️

😁 The Artbot is convincing - at least in sh*ttalking!

Just seen, grinned at and snapped in New Scientist.

So, we're through - feedback is welcome! Have a good week.

See you soon, your AInauten

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.