ChatGPT's new voice mode starts rolling out, Apple's AI gets delayed
what to know for now
🗣️ OpenAI rolls out advanced voice mode for select ChatGPT Plus users. New feature enables real-time conversations with ChatGPT, including interruptions. OpenAI delayed launch to improve content moderation and user experience. Company aims to maintain competitive edge in rapidly evolving AI market. Read more
🧠 Google's Gemma 2 2B outperforms larger AI models despite tiny size. 2.6 billion parameter model matches or exceeds GPT-3.5 and Mixtral 8x7B in benchmarks. Open-sourced, multilingual, and trained on 2 trillion tokens. Challenges notion that bigger models are always better, potentially revolutionizing on-device AI applications. Read more
🎨 Black Forest Labs launches FLUX.1, rivaling Midjourney and DALL-E 3. Startup founded by Stable Diffusion creators secures $31M seed funding. FLUX.1 offers three variants: closed-source [pro], open-weight [dev], and open-source [schnell]. 12 billion parameter models use hybrid architecture, potentially reinvigorating open-source AI ecosystem. Read more
🤖 Friend AI pendant launches, always-listening companion for $99. 21-year-old Avi Schiffmann's wearable device uses Claude 3.5 LLM, offering companionship over productivity. Experts warn of privacy concerns and potential impact on human relationships. Shipping starts January 2025, aiming to differentiate from failed AI wearables. Read more
🧪 AI Research of the Week
SAM 2: Segment Anything in Images and Videos
From MetaJake’s Take: SAM 2 (Segment Anything Model 2) is a foundation model for promptable visual segmentation in both images and videos, extending the capabilities of the original SAM to the video domain. It's basically a super-smart AI that can outline and track any object you point to in a video, thanks to its fancy memory system and a massive new video dataset they’ve created.
This model could significantly impact the AI industry by enabling more sophisticated and versatile visual understanding systems, potentially supercharging everything from self-driving cars to video editing apps with much smarter object recognition and tracking.
what to know for later
🍎 Apple delays AI rollout. Apple Intelligence launch postponed to October for bug fixes. Upgraded Siri faces further delays until late 2024/early 2025. Read more
🥊 Microsoft adds OpenAI to competitor list amid AI market rivalry. Microsoft's fiscal 2024 report now includes OpenAI among competitors in search and news advertising. Despite $13 billion investment, the companies are vying for larger share of AI market, particularly in enterprise applications and search technologies. Read more
🔍 DOJ investigates Nvidia for potential AI market dominance abuse. Probe examines allegations of bundling, strongarm tactics, and unfair allocation practices. Nvidia denies wrongdoing, emphasizing fair GPU distribution. Competitors consulted. French authorities conduct parallel investigation. Nvidia's 80% market share raises antitrust concerns. Read more
🖥️ UK shelves £1.3bn AI and tech funding promised by Conservatives. Labour government halts £800m for Edinburgh University's exascale supercomputer and £500m for AI Research Resource. Decision criticized by industry leaders, raising concerns about UK's competitiveness in crucial future industries. Read more