At Baidu World 2024, Baidu unveiled Xiaodu AI Glasses, the first native AI eyewear with a Chinese large language model. Weighing 45g, they feature a 16MP camera, 56-hour standby, quad-mic voice recognition, and functions like object recognition and translation. Launch is set for early 2025.

US Commerce bans TSMC from supplying advanced (7nm+) AI and GPU chips to China immediately. Despite restrictions, significant chip trades continue. The US plans to add ~120 Chinese firms to the restricted list, escalating tech competition.

Epoch AI's FrontierMath benchmark, featuring novel research-level math problems, shows AI models like GPT-4 struggle with advanced reasoning, achieving below 2%. Unlike previous tests with data contamination, FrontierMath highlights AI's current limitations in complex mathematical problem-solving.

DeepMind and Isomorphic Labs have open-sourced AlphaFold 3’s code and weights to academia following its creators' Nobel Prize. The model, enabling advanced protein-molecule simulations, is accessible for non-commercial research only, maintaining control over commercial applications.

Google launched Learn About, an experimental AI tool based on LearnLM, focusing on personalized education with interactive visuals. It differentiates from Gemini and ChatGPT by integrating educational research and enhancing learning experiences.

Google Gemini is introducing a standalone iOS app, beginning with a pilot in the Philippines. The app enables direct home screen access and Gemini Live voice mode, streamlining use compared to the integrated version in Google Search. Availability in other regions is upcoming.

Elon Musk’s X is broadening access to its AI chatbot Grok by offering free versions with limited queries, aiming to expand the user base and compete with models like ChatGPT. Free access, currently tested in New Zealand, requires registered accounts linked to phone numbers.

MOIRAI-MoE, developed by Salesforce AI Research, integrates Mixture of Experts into Transformers for token-level specialization. It enhances time series forecasting accuracy and efficiency, outperforming existing models in diverse and zero-shot scenarios.

Google launches Gemini, a developer-friendly AI model competing with GPT-4, integrated into the OpenAI library for seamless adoption. It offers enhanced contextual understanding and easy migration tools.

ChatGPT reached 3.7 B visits in October with strong user growth, while Google's NotebookLM surged 200% via a new podcast feature. Other AI services like Microsoft Copilot and Perplexity saw significant increases, though overall AI industry growth remains steady.

AI model compression methods—parameter pruning, quantization, and knowledge distillation—boost efficiency by minimizing model complexity and resources while preserving performance, enabling real-time applications and cost savings.

ByteDance unveils X-Portrait 2, an AI system turning photos into realistic videos with lifelike expressions using TikTok’s vast data. This innovation disrupts animation, raises misinformation issues, and supports ByteDance’s global AI growth.

Google Photos enhances video editing with AI Video Presets on Android and iOS, offering one-click effects, auto-enhance for color and stabilization, larger interface icons, and flexible speed adjustments, enabling streamlined edits without separate software.

Google launches standalone Gemini app for iOS with Gemini Live AI. Currently available in the Philippines, it provides direct home access. Broader rollout is planned after trials.

OpenAI's Orion model shows modest gains over GPT-4, reflecting AI industry stagnation as companies like Google and Anthropic also lag. To address data shortages, OpenAI uses synthetic data, while CEO Sam Altman stays optimistic about AGI.

Google launched Vids, an AI-powered app using Gemini to create video presentations from prompts. Features include stock clip insertion, script and AI voiceover generation. Available to Workspace users, supporting multiple languages with some features limited to English. Rollout began Nov 7.

AMD launches OLMo, its first open-source 1B-parameter large language model trained on Instinct MI250 GPUs. OLMo excels in reasoning and conversation, outperforming similar models. It supports deployment on AMD hardware and personal PCs, and offers open data and code.

San Francisco startup PuppyGraph, founded by ex-Google and LinkedIn staff, raised $5M for its zero-ETL graph query engine launched March 2024. It enables querying relational data as graphs without migration, reducing costs and complexity. Adopted by major firms, it enhances data analysis and AI applications.

Google launched the Utilities Gemini extension, enhancing AI capabilities and reducing reliance on Assistant. It offers voice and text controls for apps, device functions, and lock screen access to features like alarms, timers, and media playback. The rollout is ongoing, aiming for broad availability.

Google launches Google Vids, an AI-powered video creation tool integrated into Workspace. Featuring Gemini AI for automated storyboarding, customization options, and English-only voiceovers. Full AI features available until Dec 31, 2025.