Apple is set to launch a smart display next year running homeOS based on tvOS. It will support apps like Calendar and Home, feature magnetic wall mounting, and advanced smart technologies. Rumors suggest a square design with gesture control. Release details are pending.

AMD launched its first in-house compact language model, AMD-Llama-135m, on Huggingface. Featuring speculative decoding for processing 67 billion tokens efficiently, it’s open-sourced under Apache 2.0. Trained on AMD Instinct MI250 nodes, it includes a code-optimized variant for advanced NLP and coding applications.

OpenAI forecasts 2024 revenue of $3.7B with $5B losses. A $150B funding round led by Thrive Capital is oversubscribed. CTO Mira Murati departs as OpenAI plans for-profit restructuring, maintaining its non-profit arm.

OpenAI CEO Sam Altman denied rumors of significant equity stakes during a recent video meeting. The company is considering a for-profit transition while maintaining its nonprofit division. No equity compensation decisions have been made.

Apple's October keynote is set to launch new iPad and Mac models with M4 chips, including updated MacBook Pro, redesigned Mac mini, refreshed iMac, and iPad mini 6. Older models may be discontinued.

Google's Gemini AI is now integrated into Gmail with Contextual Smart Reply, offering more comprehensive response options to all users. Initially for premium accounts, the feature enhances Gmail's Smart Reply and is rolling out globally.

China Telecom AI's TeleAI team developed China’s first trillion-parameter AI model and the 115B TeleChat2-115B using fully domestic WanCard clusters. This milestone advances China’s AI self-reliance, ensuring security and efficiency through innovative training platforms and optimization strategies.

Cohere launched API V2, featuring Chat, Embed, Rerank, and Classify models with explicit versioning, enhanced Chat capabilities, improved integrations, and legacy support. Aimed at competing with OpenAI and others, this update seeks to boost developer adoption and increase market share.

Microsoft revamps Recall AI screenshot feature to be opt-in and removable. Adds encryption, Windows Hello protection, and secure processing. Preview available October for Copilot Plus PCs via Windows Insider.

OpenAI plans to raise ChatGPT's subscription from $20 to $44/month over five years to address $5B annual losses, risking user backlash as many already find $20 high.

OpenAI is set to complete a $6.5B funding round, transitioning to for-profit. Apple withdraws investment despite plans to integrate ChatGPT into Siri. The outcome will impact OpenAI’s future and AI investment trends.

OpenAI fixed a critical security flaw in ChatGPT for macOS that allowed spyware via its memory feature. Expert Johann Rehberger warned SpAIware could leak user data. OpenAI released update 1.2024.247, advising memory clearance and vigilance against suspicious links, highlighting AI privacy challenges.

A Nature study reveals advanced AI chatbots increasingly give incorrect answers without admitting mistakes. Larger models show higher error rates on complex questions, raising reliability concerns.

Google integrates Gemini AI into Workspace, enhancing tools like Gmail and Drive to foster AI agent adoption among its 8M users. This move aims to boost productivity and streamline workflows across enterprises.

Airtable launches AI tools App Library and HyperDB to enhance no-code workflows and integrate large datasets. These additions democratize AI, enabling non-developers to build custom apps and improve decision-making. A major client saved 280 hours weekly.

Google's NotebookLM now allows uploading YouTube links and audio files, enhancing its AI-powered note-taking with Gemini 1.5. The update supports diverse sources, automatic summaries, and team collaboration features, including audio overviews and public sharing, to improve research and project efficiency.

OpenAI introduced advanced voice features for ChatGPT Plus and Teams using GPT-4o, offering nine voices, enhanced accent recognition, support for 50+ languages, and personalized settings. Enterprise and education editions gain access next week, excluding some European countries. Strong privacy measures are implemented.

Hugging Face releases Hugging Chat for macOS, supporting leading open-source LLMs such as Qwen 2.5 72B and Mistral 12B. It offers web search and code syntax highlighting for free. Now in beta (v0.5.0), the company invites users to participate in testing and provide feedback.

Bilibili CEO Chen Rui unveiled "Index," a proprietary language model for AI captions, enabling real-time translation in ~10 languages with 90% accuracy. Leveraging open-source tech, it enhances multi-language content. Bilibili now has 102.3M daily and 336M monthly active users.

115 organizations signed the EU AI Convention to align with the upcoming AI Act. Despite increased interest, Commissioner Breton's resignation casts uncertainty on its future. Major tech firms show mixed participation.