E46: 7/5/24

Microsoft AI Chief says the open web is "freeware" for AI training, re-creation of human vision using AI on MRIs, Salesforce's SLM, Apple's 4M model demo and new OpenAI board seat, and more

Hope everyone had a happy and safe 4th of July! Here’s this week’s update…

PNW AI News:

• Microsoft AI CEO Mustafa Suleyman is facing major backlash after claiming publicly on CNBC that any open web content is "freeware" for AI training

• Seattle-based Tencent researchers launched a new method for synthetic data generation using AI-driven personas, to improve the creation of diverse datasets essential for training AI systems

• Amazon is planning to spend over $100B on data centers over the next decade, shifting its investment focus towards AI infrastructure

• Microsoft is hesitant to release its new AI voice model VALL-E 2, a significantly improved AI system for speech synthesis using just 2-3 seconds of sample, due to its deepfake implications

• Amazon proposes a new AI benchmark to measure retrieval augmented generation (RAG)'s accuracy

Key AI Product/Research Updates:

• Dutch researchers developed an AI system capable of reconstructing remarkably accurate images of what someone is looking at based on their brain activity recordings in an MRI (!!!!)

  • The study has limitations, but the ways this could recreate dreams, help stroke/ALS victims communicate, etc. is mind-blowing

• Salesforce has unveiled an AI model called xLAM-1B, dubbed the “Tiny Giant,” with just 1B parameters yet outperforms larger LLMs function-calling tasks

• Apple launched their 4M (Massively Multimodal Masked Modelling) AI model on Hugging Face, with the demo allowing users to generate images, perform object detection, and manipulate 3D scenes using natural language

• Cloudflare released a free tool to detect and block AI bots circumventing website scraping protections, aiming to address concerns over unauthorized data collection for AI training

• Perplexity announced new upgrades to its ‘Pro Search’ feature, enhancing capabilities for complex queries, multi-step reasoning, integration of Wolfram Alpha for math improvement, and more

• Meta announced Meta 3D Gen, a new AI system that can generate detailed, high-quality 3D assets from text descriptions in less than a minute

• WhatsApp beta revealed an upcoming 'Imagine' AI avatar generator feature, which will allow users to create personalized avatars of themselves

• Lmsys researchers published a new paper detailing RouteLLM, an open-source framework for cost-effective routing between LLMs

• RunwayML’s update, Gen-3 Alpha, adds enhancements in hyper-realistic video generation from text, image, or video inputs

• French startup Kyutai ($324M TFTD) just introduced Moshi, a new ‘real-time’ AI voice assistant capable of responding in a range of emotions and styles in a similar fashion to OpenAI’s delayed Voice Mode feature

• Researchers introduced Open-TeleVision, an open-source teleoperation system that allows users to control robots from thousands of miles away

  • Competitive to Seattle-based Olis Robotics

• A video was released of humanoid robotics company Figure AI's robots' role in potentially dangerous vehicle assembly tasks at the BMW facility in South Carolina

Key AI Business/Investment Updates:

• Apple has solidified an observer seat on OpenAI's board through a recent agreement, appointing Phil Schiller, head of Apple’s App Store and former marketing chief

• A new report from Deloitte says deepfakes will cost individuals and brands $40B by 2027 as adversarial AI gains momentum

• The Defense Advanced Research Projects Agency (DARPA) wants to use AI to find new veins of rare earth minerals to decrease reliance on imports from China

• Magic is set to raise >$200M at a $1.5B valuation, despite having no product or revenue yet, to develop its coding-specialized models that can handle large context windows

• Robinhood acquired AI investment platform Pluto, with plans to integrate advanced data analytics and personalized strategies into its trading app

PNW AI/ML/CV Fundings:

• Phaidra (Seattle, WA) raised $12M in post-Series A funding from Index Ventures and other existing investors

  • Technology that uses an array of sensors that measures various metrics and analyzes the information with software to help facilities such as data centers improve energy efficiency – TFTD $60.5M

  • CEO: Jim Gao

• Sanctuary AI (Vancouver, BC) raised an undisclosed post-Series B round by Accenture Ventures, BDC Venture Capital, and InBC Investment

  • Developer of robots designed to create human-like intelligence in general-purpose robots to achieve low-cost autonomous labor using AI – TFTD $85.82M

  • CEO: Geordie Rose

• LeyLine (Seattle, WA) raised an undisclosed Seed round from Primavera Capital Group, SternHill Associates, Taihill Ventures, Amir Husain, and Bob Stearns

  • Developer of an AI tools platform designed for animation by offering a cybernetic community facilitated by a 3D genAI-empowered tool nurturing remote collaboration among animation, VFX, and game studios – TFTD unknown

  • CEO: Yinhan L

• jigo.ai (Bellevue, WA) raised $700K in Pre-Seed funding from TheVentures, Mashup Ventures, Rabbit VC, and more

  • Developer of an automated contract diligence platform designed to provide startups with instant contract analysis by offering a full report generated within seconds and real-time AI assistance – TFTD $0.7M

  • CEO: Ryan Lee