- AInauten.net
- Posts
- π Weekly AI news: Did you miss it?!
π Weekly AI news: Did you miss it?!
π¨βπ The most important AI updates at a glance
π Weekly AI news: Did you miss it?!
π¨βπ The most important AI updates at a glance
AIHOY and happy Easter weekend, dear AInauts!
Maybe you didn't catch all the exciting news, tools and hacks about AI last week, or maybe you've just joined us. No need to worry! We'll take you down the π³ π.
We have summarized our highlights of the past week for you:
Finally, there is the most important quick news for you - so you can keep an eye on all relevant developments with just one click! Ready? Let's go!
The new ChatGPT-4o image function is causing creative explosions online! From action figure selfies to nostalgic 90s gaming scenes - everyone goes crazy. With the right prompts, you can transform yourself into a collectible action figure, have your favorite hero play PlayStation or take selfies with historical figures. Also interesting: the "Starter Packs", which summarize your digital life, and ChatGPT's cheeky analysis of your most embarrassing moments. Perfect for anyone who likes to experiment and doesn't take themselves too seriously! We have all the prompts for you, just copy and paste them.
ChatGPT dominates the app stores and is heading towards the 1 billion user mark. Sam Altman was put through the wringer in a TED Talk, while OpenAI's CFO Sarah Friar and CPO Kevin Weil also gave revealing interviews that provide particularly exciting insights. Absolutely worth seeing!
π©βπ¨ Better design with AI help: New Canva updates
Canva presented really cool AI features at the Create 2025 conference! With one click, users can magically edit images, generate backgrounds and adjust colors like professional designers. Very helpful: AI-supported voiceovers, karaoke-style subtitles for videos and even a new sheets function for spreadsheets. Plus: You'll soon be able to create code and apps with Canva AI!
Google is also presenting a wave of exciting AI innovations! The new AI function in Google Sheets enables text generation using a simple formula. Simply enter =AI(...). Media tools such as Chirp3 for voices and Lyria for music are introduced, and the free Veo2 for video generation is already widely available.
Further highlights: The release of Firebase Studio for coding, the powerful Dragontail model (still "secret"), PDF-to-Mindmap in NotebookLM and audio summaries in Google Docs. The Agent2Agent protocol was also introduced and a 68-page prompt engineering whitepaper was published.
OpenAI has introduced several new models this week. The new GPT-4.1 offers a gigantic context window of 1 million tokens, is fast and inexpensive, but is currently only available via API. The newly presented reasoning models o3 and o4-mini are also particularly exciting: they can master complex mathematical, scientific and visual tasks and combine all ChatGPT tools independently. O3 is the most powerful reasoning model to date. Developers can also use the new Codex CLI for terminal-based coding.
Google has published a comprehensive guide with prompting techniques for AI language models. The most important methods are one-shot/few-shot prompting with examples, role prompting for targeted answers and precise instructions instead of restrictions. Advanced techniques include Chain-of-Thought for complex problems, Step-back Prompting for broader knowledge and Tree-of-Thoughts for creative tasks. For best results: Reasoning LLMs for complex tasks, use regular models for simple queries.
With version 2.0, KLING AI shows impressive progress in video and image generation. The new master model processes complex sequences with smoother movements, while KOLORS 2.0 can handle over 60 different image styles. The absolute highlight, however, is the new multimodal editor: it allows you to seamlessly combine and edit videos and images. For example, you can easily insert objects into existing videos - like a cute stuffed animal Nauti in a gift box. It's pretty impressive what this AI can do!
π° AI news quickie: HAI highlights from the industry
The last few days have once again been packed with exciting AI news. Ready? Let's go!
OpenAI updates in a multipack
OpenAI's new models GPT 4.1 (prompting tips here), o3 and o4-mini are already being enthusiastically tested, used for trading and critically scrutinized by many users.
OpenAI has also published the BrowseComp benchmark, which tests the performance of AI agents in finding hard-to-access information - over 1,200 tricky questions are waiting to be cracked!
According to Sam Altman, OpenAI has almost 800 million users, almost 10% of the world's population! And the growth is explosive, thanks to image generation.
ChatGPT was the world's most downloaded app (excluding games) in March, beating Instagram and TikTok.
ChatGPT now also has an image library for your AI artwork!
A new trend is circulating: Users are using ChatGPT to track locations on photos ...
Now the AI giant is also planning its own social network to compete directly with Elon and Zuck.
OpenAI has published a number of practical guides to help companies and developers drive AI implementation forward.
The Pioneers Program is also intended to support companies in the practical use of AI in sectors such as healthcare and finance.
In order to gain access to the most advanced models in the future, you will probably have to be verified...
The $500 billion infrastructure project "Stargate" is also considering investments in the UK, Germany and France.
OpenAI is planning to buy Windsurf for around $3 billion, after rival Cursor was first mooted. It is a deal that would be the company's largest acquisition to date.
Google and Gemini
π¬ Google has developed DolphinGemma, a fascinating AI model to help decode the communicative diversity of dolphins!
Gemini Advanced subscribers can try out the new AI video tool Veo 2 to create realistic videos with just one prompt.
Deep Research is also now available in Gemini 2.5 Pro Experimental
Screen sharing feature now available for free for all Android users
Google Gemini can now answer specific questions about Google Photos.
Google AI Studio has been given a fresh coat of paint: Starter apps, improved design and a user-friendly dashboard for developers.
Google Classroom has introduced a new AI function that helps teachers to generate questions. Simply enter texts or upload files.
Google has blocked 39.2 million ad accounts for fraud with the help of AI - three times as many as last year.
Claude, Copilot, Grok
Claude gets a cool research feature (initially only available for the expensive subscriptions) and the connection to Google Workspace - an ideal combination. And with an MCP connection, you can expand Claude's memory .
A new study shows that students mainly use Claude to create learning materials and solve problems, especially in math and computer science.
What can Microsoft's Copilot actually do? Quite a lot, try it out.
Microsoft has introduced a new "computer use" for Copilot Studio that allows AI agents to work interactively with websites and apps!
Grok now has a Canvas-like tool called Grok Studio for document and app creation - as well as Google Drive integration!
A developer created a test with SpeechMap to see how well AI chatbots handle controversial topics. Conclusion: Grok 3 has few limits.
AI tools and startups
A resourceful "hacker" has teased out the full system prompts of FULL v0, Manus, Cursor, Same.dev, Lovable & Devin AI...
GeoSpy turns image data into precise location information - without EXIF ...
QRCraft turns your QR codes into mini works of art.
AI Product Rankings lets you discover how the hottest AI models push certain products and brands - and how you can optimize for them!
With MCPify.ai you can build MCP servers in record time, no coding required! Ideal for task scheduling, social media and much more.
Notion has introduced Notion Mail, an AI email client for Gmail.
Trends and benchmark data
The Stanford AI Index presents the AI situation in 2025 - investment, public opinion and performance in detail.
The Forbes 2025 AI 50 list is here! New stars like Anysphere and Mercor hold their own in the AI world alongside giants like OpenAI and Anthropic.
Scale AI co-founder Lucy Guo became the youngest self-made billionaire at the age of 30.
LiveBench is a benchmark for LLM that continuously simplifies the objective evaluation of models.
A new analysis shows the price and LiveBench performance of various models. Google dominates the price-performance ratio.
KPMG's AI bot Kai has reduced interview scheduling time and saved over 1,000 hours for the recruitment team!
Research, safety and ethical concerns
Demis Hassabis, CEO of Google DeepMind and Nobel laureate, sees AGI in 5 to 10 years - and warns of the ethical risks it poses.
Wow! One-minute videos are now generated automatically - consistent Tom and Jerry cartoons are the first proof that it works.
Biotech startup Profluent has discovered "scaling laws" that improve the performance of AI models in protein design - a major milestone!
Your AI digital copy can now take over meetings and provide comfort.
Wikipedia partners with Kaggle to provide an optimized dataset to stop AI developers from scraping.
Meta now trains AI in the EU with public content to better understand cultures and languages, while Ireland's data protection authority scrutinizes X's handling of European user data.
And: the European General Data Protection Regulation (GDPR) could be reconsidered by the summer - say what?!
The US Marines are now using generative AI for surveillance... but the US is lagging behind China in military biotech dominance, and a new report warns that Washington must act now.
The US authorities are getting cold feet and taking action against DeepSeek and Nvidia. Nvidia now needs a license to export to China, which could lead to an expected loss of 5.5 billion dollars.
Unitree has launched the RobOlympics. Viewers will soon be able to experience the duels between π¨π³ and πΊπΈβ¦
And thatβs a wrap! Enjoy the rest of your weekend.
Your AInauts, And thatβs a wrap!
Fabian & Reto
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