E64: 11/8/24
Meta and Anthropic go after US Gov/Defense use cases with Palantir, AIexa AI is delayed once again, Python popularity with AI, Saudi Arabia's $100B initiative, and more...
Happy Friday, from Nashville! Here’s this week’s busy update…
PNW AI News:
Microsoft teased that its ‘Copilot Vision’ feature is coming ‘very soon,’ enabling the AI assistant to see and understand a user’s browser content and behavior
What lightweight VLM is Microsoft using for this in-browser vision??
Amazon has reportedly delayed the rollout of its AI-infused Alexa to 2025, as testing has faced technical challenges, including hallucinations and deteriorating performance on basic tasks
Microsoft researchers introduced Magnetic-One, an AI orchestration system that coordinates multiple specialized agents to tackle complex real-world tasks like writing code, operating a browser, and even ordering food from a restaurant
Amazon Prime Video released X-Ray Recaps, a new AI-powered feature that generates summaries at any viewing point to keep viewers up to date without spoilers
T-Mobile will pay $100M to OpenAI over the next three years to develop an ‘intent-driven’ AI platform that can take actions for users and integrate with operations and transaction systems for customer service tasks
Amazon secured FAA approval to deploy its new computer vision-guided MK30 delivery drones, enabling beyond-line-of-sight flights and moving the company closer to broader autonomous deliveries
Key AI Product/Research Updates:
Python has become the most-used programming language on GitHub, largely accredited to the AI surge
After users briefly accessed the new OpenAI o1 model via "chatgpt.com/?model=o1" before it was restricted, the LLM reportedly can handle images in prompts, and can process up to 200K tokens
AI labs Decart and Etched launched Oasis, an AI model that generates playable video game environments in real-time, alongside a playable Minecraft-style demo as a demonstration
Operates at 20 FPS, 100x faster than traditional AI video generation models, even with the added complexity of user interaction
Runway unveiled Advanced Camera Control for its Gen-3 Alpha Turbo model, bringing new precision to AI-generated video outputs with features that mirror traditional filmmaking techniques
Anthropic released PDF support for its Claude 3.5 Sonnet model in public beta, unlocking the ability to analyze both text and visual documents up to 32MB like charts and images
Tencent just released Hunyuan-Large, a new open-source LLM that combines scale with a Mixture-of-Experts (MoE) architecture that matches LLaMa-405B in performance
Apple’s iOS 18.2 Beta 2 revealed that ChatGPT integration with Siri will include daily usage limits for free users and a $19.99 monthly Plus upgrade option, offering expanded access to GPT-4o features and DALL-E image generation
Apple is reportedly taking its first steps toward potential smart glasses development with a new internal research initiative called ‘Atlas’
Key AI Business/Investment Updates:
Saudi Arabia unveiled plans for "Project Transcendence," a $100B AI initiative to establish the kingdom as a global tech powerhouse through investments in data centers, startups, and infrastructure
Meta will make its LLaMa AI models available to US government agencies and defense contractors (citing Amazon, Microsoft, Palantir, Lockheed, Oracle, and more) for national security applications
Marks a stark shift in the company's policy around the military use of its open-source AI…
Earlier in the week… Chinese military researchers reportedly used Meta's open-source LLaMa-3.1 model to develop ChatBIT, an AI tool designed for military intelligence analysis and strategic planning
Meta's terms prohibit use for "Military, warfare, nuclear industries or applications, espionage", etc. but has limited ways to enforce this
Anthropic just announced a major partnership with Palantir and AWS to bring its Claude AI models to US intelligence and defense agencies
Following suit of OpenAI/Meta
Perplexity CEO Aravind Srinivas responded to the New York Times’ Tech Guild strike, offering AI help with election coverage
Comes right after Perplexity announcing a new dedicated hub for U.S. election news
AI startup Physical Intelligence just secured $400M in funding at a $2.4B valuation led by Jeff Bezos and OpenAI, as it unveils its π0 (pi-zero) model for general-purpose robot control
OpenAI just acquired the $15M+ domain name chat.com (which now leads to ChatGPT) from HubSpot's founder
One of the largest domain deals in history
Could ChatGPT become just "Chat" sans GPT???
Disney launched a new ‘Office of Technology Enablement’ group responsible for managing AI and mixed reality adoption within the company
Humanoid robot artist Ai-da just made history at Sotheby’s Auction House, with its portrait of computer science pioneer Alan Turning selling for $1.3M, marking the first major auction sale of a robot-created artwork
PNW AI/ML/CV Fundings:
WorkHelix (Beaverton, OR) raised $15.33m of seed funding from undisclosed investors
Developer of generative AI designed to create an interactive, dynamic roadmap tailored for the organization – TFTD $15.33m
CEO: James Milin
Logic (Seattle, WA) raised $4.3M in Seed funding from Founders’ Co-op, Audacious Ventures, Ali Partovi’s Neo, several current and former Brex execs, Convoy co-founder Dan Lewis, and more
Turns business docs in natural language into executable APIs with AI – TFTD $4.3M
CEO: Steve Krenzel