Check out past editions of the newsletter by clicking here

Hey AI team! We’re back with a new edition of Five Minute AI this Friday ⏰

On today’s menu:

• 1X Technologies’ NEO Beta humanoid robot promises to be a real help with household chores

• Amazon supercharges Alexa with Anthropic’s Claude AI for richer conversations and better smart home control

• OpenAI Japan reveals details about GPT-Next, promising performance 100 times better than GPT-4

• Visual Electric upgrades its AI image generator with 55 different graphic styles

Ready? Let’s go...

TOP STORY

🤖 NEO Beta: the humanoid robot that’s going to change your daily life

The race for home robots just hit a new milestone with 1X Technologies’ announcement of NEO Beta. It’s a strikingly lifelike humanoid robot designed to help out with everyday household tasks. What sets this new robotic assistant apart are its advanced capabilities in movement, interaction and task execution.

Details:

• NEO Beta stands 1.65 m tall, weighs 30 kg, and can walk at 4 km/h or run at 12 km/h

• It can carry up to 20 kg and run for 2 to 4 hours on a single charge

• The robot uses embodied artificial intelligence to understand its surroundings and learn from past interactions

• It’s designed to handle a wide range of tasks, from household chores to assisting people with limited mobility

Why it matters

NEO Beta marks a significant step forward in the race for an affordable home robot. With this launch, 1X Technologies takes its place alongside major players like China’s AGIBOT fleet and Tesla’s Optimus. Embodied AI lets NEO understand its environment and learn intuitively, paving the way for more natural interactions in our daily lives. It’s so lifelike that plenty of people on social media even argued over whether it was a person in a costume!

🗣️ Amazon Alexa levels up with Claude AI!

Image made with Visual Electric, Flat Design style.

Amazon is getting ready to launch a new version of Alexa, this time powered by Claude, the AI built by the startup Anthropic. This major upgrade promises to radically improve the user experience.

Details:

• Early versions of Alexa, built on Amazon’s own AI, suffered from slow response times and disappointing performance.

• So Amazon went with Claude AI, an unusual move for a giant that usually prefers its own technology.

• The new Alexa will offer richer conversations, personalized shopping advice and better smart home control.

• Amazon is considering charging between $5 and $10 a month for this upgraded version, while the classic version will stay free.

Why it matters:

Even tech giants like Amazon run into trouble building high-performing AI systems. This strategic about-face, leaning on Claude instead of its own technology, shows just how complex the AI race is. While Google is building its own model and Apple is working with OpenAI, Amazon is betting on outside expertise. The future of voice assistants is shaping up to be exciting!

🚀 OpenAI Japan reveals details on ‘GPT-Next’

I mentioned it in a previous newsletter: OpenAI is preparing to release its new GPT model, and its Japanese division just spilled some juicy details about GPT-Next, which promises to be 100 times more powerful than GPT-4.

Details:

• GPT-Next would reportedly deliver a 100x performance gain while using minimal computing resources.

• This technological leap comes from a better architecture and greater learning efficiency, rather than a brute-force increase in computing power.

• GPT-Next will use a miniaturized version of “Strawberry,” an advanced system currently in development at OpenAI.

• Rumors suggest the model could launch by the end of 2024, though that remains to be confirmed.

Why it matters:

OpenAI remains the undisputed leader in AI model usage. If the information is accurate, GPT-Next could be the next big leap users have been waiting for. The most impressive part is that this advance would come without a significant increase in computing needs, a major plus for applications that require speed, efficiency and lower costs. An even smarter model could push its uses into even more complex areas and give us a foothold in a world we’ve so far only seen in science fiction movies.

LEARN TO MASTER AI

🎨 Master AI with Visual Electric

Visual Electric recently upgraded its image generator with version 2 of its model. It doesn’t just deliver better renders now; it also offers a wide variety of graphic styles, perfect for all kinds of creative needs.

Tutorial steps:

1. Select a style: The current model lets you choose from 55 different styles, organized into 5 main categories: Photography, Fine art, Illustration, Crafts and Digital. Each style comes with visual samples to guide you.

The different styles Visual Electric offers

2. Enter your prompt: For example, if you want to create a warm atmosphere for an interior design project, use a prompt like:

spacious and modern living room with light wood furniture, plants, and large windows

3. Run the generation: You’ll get four image variations for each selected style, so you can pick the one that best matches what you’re looking for.

Generations from the same prompt in several different styles

4. Use the visual editor: Visual Electric offers a very intuitive editor where you can move, enlarge or adjust images as needed. It’s a truly flexible and easy-to-use workspace.

Their free plan lets you generate up to 40 images a month, which is already more than enough to explore the platform. If you need more, there’s a $20/month subscription that gives you unlimited generations.

That makes Visual Electric a versatile, accessible tool, ideal for creatives who want to explore different artistic directions without compromising on quality.

WHAT THE RESEARCH SAYS

🧬 OpenCRISPR-1: AI revolutionizes gene editing

Researchers at Profluent Bio have hit a major milestone by creating OpenCRISPR-1, a new gene-editing tool designed by artificial intelligence. This innovation promises performance equal to or better than current CRISPR systems, while offering improved safety.

Details:

• Uses large language models (LLMs) trained on more than a million CRISPR operons to generate new protein structures

• OpenCRISPR-1 proves just as effective as current models at cutting DNA, with fewer errors

• The new tool can handle a variety of tasks, including base editing

• The OpenCRISPR-1 sequence has been made public to encourage its use in research

Why it matters

This breakthrough paves the way for safer, faster and more personalized treatments for genetic diseases. Although practical applications are still a long way off, OpenCRISPR-1 hints at the possibility of creating custom gene editors tailored to each individual’s DNA. That could dramatically speed up the development of treatments for diseases that were previously incurable. To dig deeper, I recommend this article in French, which offers a detailed analysis of what this technology means.

🖥️ DisTrO: AI within reach of your personal computer

Nous Research just unveiled DisTrO, a new method for training large AI models that dramatically reduces the need for an ultra-fast internet connection. This innovation could make it possible to train AI on standard home computers.

Details:

• Cuts the amount of data shared between computers by 857 to 3,000 times during pre-training, and by up to 10,000 times during fine-tuning

• The method is compatible with various model types and network setups

• Successful tests training a 1.2 billion-parameter language model, with performance on par with traditional methods

• Potential for decentralized AI training you can do from home

Why it matters

Right now, training cutting-edge AI models is the preserve of big tech companies with massive computing resources. DisTrO could democratize the process by letting individuals and small organizations take part in AI development using standard computers and internet connections. This breakthrough opens the door to greater diversity in AI development, potentially boosting innovation and making the technology more accessible.

THE LATEST TRENDING TOOLS

AI language models (LLMs) get updated so often that it’s hard to know which ones are the most competitive. The site Artificial Analysis promises a constantly updated, independent evaluation of the various LLMs across several criteria (quality, speed, price, etc.).

💡 OTHER TOOLS:

• 📌 PackPack revolutionizes how you manage bookmarks and online content, using AI to organize your important information and find it easily.

• 🕺 Viggle brings characters to life with AI motion transfer, opening up new possibilities for animation and visual content creation.

• 📊 GoMarble Ad Analyzer digs deep into ads to deliver valuable insights on their performance, helping marketers optimize their campaigns.

• 🎤 SpeakHints acts as a real-time, AI-powered speaking copilot, improving your communication skills in all kinds of situations.

• 📝 Mimrr automates documentation and analysis for development teams, making it easier to track and optimize software projects.

• 🤖 AgentOps provides an efficient environment for building and debugging AI agents, speeding up the creation of high-performing AI.

• 🎨 Krea FLUX Style Mixer lets you combine multiple Flux image generation styles with full control, sparking digital artists’ creativity.

• 📊 Next Alpha Andromeda bills itself as an LLM specialized in crypto research, offering in-depth analysis for investors and enthusiasts.

IN THE NEWS

• Brazil temporarily suspends access to X (formerly Twitter) over court violations tied to the fight against disinformation, sparking debate over free speech and social media regulation.

• Waymo hits a major milestone with more than 100,000 paid rides a week in autonomous vehicles, showing the progress of its driverless technology in several US cities.

• Playground introduces a new AI-powered graphic design tool that lets users intuitively create logos, social media visuals and t-shirt designs for free.

• Fourier Intelligence unveils GR-2, a next-generation humanoid robot featuring innovations like swappable batteries and improved manual dexterity.

• A London school is set to open the UK’s first teacherless classroom, using AI platforms to deliver personalized GCSE instruction to students.

• YouTube rolls out a new policy requiring creators to disclose AI-generated content in their videos, especially for voice cloning or face swapping.

• Bland AI offers a platform that lets businesses automate their phone operations with artificial intelligence, streamlining call handling.

• Amazon beefs up its robotics capabilities by hiring experts from Covariant AI and licensing their foundation models, aiming to improve warehouse automation.

• Stanford researchers Medicine have developed nuclei.io, an AI tool designed to help pathologists identify diseased cells faster and more accurately, promising advances in medical diagnosis.

YOUR DAILY CONTENT FIX

1/ 🤖 A fascinating look at the NEO beta robot and the ambitions of the company 1X (18 min, in English)

They Built This Robot For Your Home | 1X Technologies

2/ 🌍 A captivating exploration of AI’s potential societal impacts on jobs (14 min, in English)

If AI Takes All The Jobs, This Is What Society Looks Like

3/ 🎵 MC Solaar breaks new ground with a fully AI-generated music video for his new album, blending human creativity and technology

MC★Solaar – Tout est vrai (Lyrics visualizer)

THAT’S ALL FOR TODAY

Thanks for reading the work of our purr-sistent team 😼

💡 Help me improve or come up with new ideas → [email protected] or @mattioo81

👍️ Like what you’re reading? Subscribe!