Apple's AI model revealed!

PLUS: xAI officially open-sources Grok

Welcome, AI enthusiasts.

Apple's AI efforts have often been shrouded in secrecy, but the company just pulled back the curtain with the release of MM1.

The powerful new AI model could indicate a new era of AI for Apple — and be the key to unlocking Siri's true potential. Let’s explore…

In today’s AI rundown:

  • Apple unveils new MM1 AI models

  • xAI officially open-sources Grok-1

  • Instantly translate videos with AI-powered lip sync

  • Google’s VLOGGER brings avatars to life

  • 6 new AI tools & 4 new AI jobs

  • More AI & tech news

Read time: 4 minutes

LATEST DEVELOPMENTS

APPLE

Image source: Apple

The Rundown: Apple researchers just published a new paper unveiling MM1, a family of multimodal AI models that combine visual and language understanding to enable advanced capabilities.

The details:

  • MM1 models were trained on a carefully curated mix of image captions, image-text data, and text-only data.

  • The largest 30B parameter model showed a strong ability to learn from only a handful of examples and reason over multiple images.

  • The research found that scaling the model’s image processing had the biggest impact on performance.

  • MM1’s benchmarks compete with state-of-the-art multimodal models like GPT-4V and Gemini Pro.

Why it matters: Apple’s level of detail and lack of fanfare for this model release is a big departure from its typical secrecy — and a massive win for open source. With a capable model now officially a reality, is it finally time for Siri to level up?

TOGETHER WITH DENTORO

The Rundown: Dentoro, a software engineering company founded by ex-Googlers and Harvard grads, can help you build an MVP or develop custom AI solutions at unbeatable prices.

With Dentoro, you can:

  • Fast-track your AI project with MVPs starting at just $36/hour.

  • Benefit from tailored AI solutions to address specific business challenges and objectives.

  • Easily scale AI initiatives from prototype to production.

Take advantage of Dentoro's expertise to build your next AI prototype. Get started now and enjoy 30% off your first month.

XAI

Image source: xAI

The Rundown: Elon Musk and xAI just released the weights and architecture of its massive 314B parameter language model, Grok-1, under an open-source Apache 2.0 license.

The details:

  • Grok-1 is a Mixture-of-Experts model, with only 25% of its weights active for any given input token to enable more efficient computation.

  • The released model is the raw, pre-trained checkpoint from October 2023, not fine-tuned on any specific tasks.

  • xAI provided instructions on its GitHub repo for developers to get started, also publishing the model on Hugging Face.

Why it matters: By open-sourcing one of the world's largest LLMs, xAI is walking the walk in Musk’s perceived moral battle against OpenAI’s closed models. While Grok’s capabilities don’t break any new barriers yet, the move is another big win for collaborative and transparent AI development.

AI TRAINING

The Rundown: Pipio's AI-driven video translation tool allows you to easily translate your videos into multiple languages while preserving the original voice and automatically syncing lip movements.

Step-by-step:

  • Create your account on Pipio's website.

  • Select and upload your video to the platform, and pick the target language for translation.

  • Click the submit button and let Pipio's AI handle the translation.

  • Export your translated video for further editing, or share it directly on social media!

AI RESEARCH

Image source: Google

The Rundown: Google researchers just developed VLOGGER, a new AI model that can generate photorealistic talking avatar videos with full upper body motion from just a still image and audio clip.

The details:

  • VLOGGER creates a controllable avatar that captures likenesses and movements.

  • The model was trained on a large multimedia dataset containing 800,000 videos of people talking with labels for each part of the face and body.

  • Potential applications include dubbing videos in other languages, creating realistic avatars for games or assistants, and enabling low-bandwidth video chats.

Why it matters: Whether it's enabling realism for AI assistants, allowing real-time video dubbing across languages, or letting us video chat as our favorite avatars — models like VLOGGER point to a future where the boundaries between our physical and digital selves blur in captivating new ways.

NEW TOOLS & JOBS

  •  Cykel AI- Automates tasks across the web. Works with your favorite tools. Try for free.*

  • 📄 Abel- AI-powered document review for law firms

  • 🎬 Katalist- Generate visual stories with consistent characters from a script

  • 💼 Octolane AI- Real-time B2B data enrichment for businesses

  • 🤖 Synthflow AI- Create conversational AI voice agents without coding

  • 📝 Zeda Co-Pilot- Extracts customer feedback and turns it into product insights

  • 🔬 Spotter- Senior Applied Data Scientist

  • 💻 OpenAI - Backend Engineer, Sora

  • 📚 Cohere - Data Annotator - Advanced Math

  • ⚖️ Mistral AI - Legal Counsel

*Sponsored listing

QUICK HITS

India’s Ministry of Electronics and IT revised its AI advisory to remove the requirement for firms to gain government approval before deploying AI models, following backlash from global entrepreneurs and investors.

Cognition’s Devin showed off its autonomous AI agent skills in a viral X post, with the model opening a support ticket and exchanging Slack messages to solve a coding issue on its own.

Senator Bernie Sanders proposed a bill establishing a four-day work week in the U.S. with no loss of pay, citing productivity gains from AI and automation.

Maisa announced the beta release of its Knowledge Processing Unit, which sets new standards in reasoning, comprehension, and problem-solving by combining the power of LLMs with decoupled reasoning and data processing.

OpenAI researcher Leopold Aschenbrenner posted that the past year since GPT-4’s release will be ‘the slowest 12 months of AI progress for quite some time to come’.

Rapper Tyler The Creator said in an interview that he is not afraid of the rise in AI, saying the tech will never catch up to his creative abilities.

Reddit is facing an investigation from the FTC over its licensing of user data for AI training, revealed in forms filed to the SEC as the company prepares for its IPO.

THAT’S A WRAP

SPONSOR US

Get your product in front of over 500k+ AI enthusiasts

Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners around the world. Get in touch today.

FEEDBACK

How would you rate today's newsletter?

Vote below to help us improve the newsletter for you.

Login or Subscribe to participate in polls.

If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.

Join the conversation

or to participate.