Use AI as Your Personal Assistant

Ready to embrace a new era of task delegation?

HubSpot’s highly anticipated AI Task Delegation Playbook is your key to supercharging your productivity and saving precious time.

Learn how to integrate AI technology into your processes, allowing you to optimize resource allocation and maximize output with precision and ease.

Download Now!

Today:

• Sam Altman Never Before Seen Footage Interview

• Apple Exits OpenAI Funding Talks

• Google DeepMind's New AlphaChip 

• Google Adds Gemini to Workspace

• Cohere Updates AI Tools Suite

Sam Altman SHOCKING "Catalysts of Change" Interview (never before seen footage)

Sam Altman, CEO of OpenAI, announced that Mira Murati, the Chief Technology Officer, and Bob, the head of research, are leaving the company. He expressed gratitude for their contributions and is eager to work with new technical leaders.

Altman discussed OpenAI's rapid growth and new products like Voice Mode, which allows people to talk naturally with AI, and improved models that think more logically. He emphasized the importance of innovation and plentiful energy in making lives better. Altman encouraged entrepreneurs, especially in Europe and Italy, to seize this unique moment to build great companies and drive technological progress.

WATCH THE VIDEO ON YOUTUBE

Apple Exits OpenAI Funding Talks

Apple has pulled out of talks to invest in OpenAI's funding round, which aims to raise $6.5 billion. This move ends what could have been a rare investment by Apple in another major tech company. Microsoft and Nvidia are still in discussions, with Microsoft expected to contribute about $1 billion, adding to the $13 billion it has already invested. Venture-capital firm Thrive Capital is leading the round with a $1 billion investment. 

OpenAI is also transitioning from a nonprofit to a for-profit company—a complex change encouraged by investors. If not completed in two years, current investors can ask for their money back.

Google DeepMind's New AlphaChip

AlphaChip is an artificial intelligence method developed by Google DeepMind to design computer chip layouts faster and better than humans. It uses reinforcement learning, a technique where AI learns by trial and error, treating chip design like a game to place components efficiently. This reduces design time from months to hours and creates layouts that outperform human designs. 

AlphaChip has been used to create Google's Tensor Processing Units (TPUs), specialized chips that power AI systems like large language models. Companies like MediaTek are also adopting it. AlphaChip is transforming chip design, leading to faster, cheaper, and more energy-efficient chips.

Google Adds Gemini to Workspace

Google is adding its Gemini artificial intelligence features to Workspace, aiming to bring smart helpers to millions of users. Workspace includes apps like Gmail and Google Drive, used by over 8 million paying customers. Gemini will assist with tasks like summarizing emails, finding information in documents, and creating charts from simple requests. 

Starting in October, users can set up automatic actions where Gemini reads and sorts emails or documents and takes steps like sending invoices to the right team. Google believes this will boost productivity by handling tedious tasks, making businesses more competitive without needing extra help.

Cohere Updates AI Tools Suite

Cohere has updated its AI tools—Chat, Embed, Rerank, and Classify—to help developers switch from competitors like OpenAI. The new version, called API V2 (a set of rules that let software programs talk to each other), now requires specifying the model version to keep performance consistent. 

Inputs are simplified into one messages array, allowing for more complex uses. Tools are defined using JSON schema—a flexible way to describe data—making them compatible with more applications. Streaming now uses Server Sent Events (SSE), which improves responsiveness by sending updates in real-time. Cohere aims to attract more developers by making its platform easier to use than larger AI companies'.

🧠RESEARCH

MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models

MaskLLM reduces the computational load of large language models by learning which weights to keep and which to discard. It does this by selecting a fraction of connections in each group to retain, making the model smaller and faster without much loss in performance. This method works well across different tasks and models.

LLaVA-3D: A Simple yet Effective Pathway to Empowering LMMs with 3D-awareness

LLaVA-3D is a framework that enables large models to understand 3D scenes in addition to 2D images. It introduces "3D Patches," which connect features from 2D images to their positions in 3D space. This approach trains faster and achieves top results in 3D tasks without losing 2D capabilities.

MIMO: Controllable Character Video Synthesis with Spatial Decomposed Modeling

MIMO is a new method to generate realistic character videos where users can control the character, motion, and scene. They transform 2D videos into 3D by estimating depth and separating layers (character, background, moving objects), encoding them to guide video creation. This allows flexible control and realistic scene interactions.

A Preliminary Study of o1 in Medicine: Are We Closer to an AI Doctor?

Researchers assessed OpenAI's new language model, o1, to see if we're closer to having an AI doctor. They tested its understanding, reasoning, and ability to work in multiple languages across 37 medical datasets, including tough quizzes from top medical journals. o1 outperformed previous models like GPT-4 but still has flaws like making up information and uneven performance in different languages.

YesBut: A High-Quality Annotated Multimodal Dataset for evaluating Satire Comprehension capability of Vision-Language Models

"YesBut" is a dataset of satirical and regular images to test if AI models can understand humor and satire. The dataset includes tasks like detecting satire and explaining why an image is funny. They found that current models struggle, showing a gap in AI's ability to grasp humor.

🛠️TOP TOOLS

Polymet - AI Product Designer

Arcade - Turn your thoughts into things

HeyGen - Create and translate videos with AI

Nuelink - Helps you organize, automate, analyze and manage your social media from one place and saves you time to focus on your business while your social media runs itself.

OpenMusic - A next-gen diffusion model designed to generate high-quality music audio from text descriptions

📲SOCIAL MEDIA

🗞️MORE NEWS

• Apple might launch a smart home display in 2025, running homeOS—a new system based on Apple TV's software. It would support apps like Calendar and Notes and could be mounted on walls using magnets.

• Spotify expanded its AI Playlist feature to Premium users in the U.S., Canada, Ireland, and New Zealand. This tool lets users create custom music playlists by typing in prompts, helping them discover new songs easily

• Figma relaunched its AI tool 'First Draft' after fixing issues where it copied designs. 'First Draft' helps designers quickly create app layouts using artificial intelligence, offering templates from simple sketches to detailed designs.

• Warner Bros. Discovery is partnering with Google to use AI to create captions for their shows more quickly and cheaply. Human workers will still check the captions to make sure they're correct.

• Archaeologists used AI and drones to discover 303 new ground drawings near Peru's Nazca Lines, nearly doubling known figures. Dating to 200 BC, these small images of animals and people help understand ancient cultures.

• Harvard scientists developed an AI tool called TxGNN that discovers new uses for existing drugs to treat rare diseases. It finds possible treatments and predicts side effects, offering hope for conditions without medicines.