RAIZOR Report: OpenAI’s Orion Faces Hurdles, TSMC Restriction’s Impact, AI in Music, Robotics, and Content Creation

OpenAI facing challenges with new model Orion

OpenAI employees testing the new AI model ‘Orion’ have reported that while its overall performance is better than existing models, the rate of improvement is lower than in past upgrades. The main issue is a lack of new, high-quality, diverse training data. OpenAI has formed a foundations team to overcome this, trialing synthetic data alongside real-world data. They’re also planning to use techniques like reinforcement learning and fine-tuning on specific tasks post-training to address performance gaps. This development could impact businesses relying on OpenAI’s technologies, potentially slowing down the pace of AI advancements in various applications.

TSMC banned from exporting AI chips

The United States Department of Commerce has imposed restrictions on Taiwan Semiconductor Manufacturing (TSMC), preventing them from exporting advanced AI chips to China. This decision comes after discovering TSMC AI chips were integrated into one of China’s most advanced AI chips, produced by Huawei. The restrictions are set to take effect from November 11, and TSMC has already informed its Chinese customers. This move could create major disruption for China’s advancing AI industry and impact plans of companies like Baidu and Alibaba to build state-of-the-art language models. Businesses relying on Chinese AI technologies may need to reassess their strategies and supply chains.

Salesforce’s hiring spree to sell AI agent

Salesforce is planning to hire over 1,000 employees to sell its new AI agent tool called Agentforce. Launched last month, Agentforce enables companies to deploy autonomous agents to complete tasks without human intervention. It can perform multiple roles and enhance employee productivity by responding to changes in data, business rules, automation, or signals from external systems. This development could significantly impact businesses looking to streamline operations and improve efficiency through AI-powered automation.

OpenAI loses another key safety exec

OpenAI has lost another key safety researcher, Lilian Weng, who was Vice President of Research and Safety, after working at the start-up for over seven years. This departure comes after a long line of AI safety researchers and executives have left the company, including former members of the disbanded superalignment safety team, Ilya Sutskever and Jan Leike. The ongoing exodus of safety experts from OpenAI could raise concerns for businesses relying on their AI technologies, potentially impacting the development and implementation of safe AI systems.

Altman predicts AGI in 2025

OpenAI CEO Sam Altman has predicted that artificial general intelligence (AGI) will be achieved in 2025. In an interview with Y Combinator founder Gary Tan, Altman stated that the path to AGI is ‘basically clear’ and will require engineering, not new scientific breakthroughs. This prediction comes alongside conflicting reports of slowing progress in large language model development and scaling across the industry. Businesses should closely monitor these developments, as the potential arrival of AGI could have far-reaching implications across various sectors and industries.

The Beatles make AI history with Grammy noms

“Now and Then,” The Beatles’ AI-enhanced final song, has become the first AI-assisted track to receive Grammy nominations, marking a historical moment for AI’s role in music production. The song earned nominations for Record of the Year and Best Rock Performance. The track used AI “stem separation” technology to clean up and isolate John Lennon’s vocals from a 1978 unreleased demo. This development showcases the potential for AI in the music industry, opening up new possibilities for businesses in audio production, restoration, and creative content generation.

MIT’s AI trains robot dogs in virtual worlds

Massachusetts Institute of Technology researchers have unveiled an AI system called LucidSim that trains four-legged robots using generated imagery, achieving unprecedented real-world performance without ever seeing actual environments during training. LucidSim combines physics simulations with AI-generated scenes to create diverse training environments for robotic learning. Robots trained in LucidSim’s artificial environments completed complex tasks like obstacle navigation and ball chasing with up to 88% accuracy. This advancement could revolutionize robotics training for businesses, potentially reducing costs and improving the efficiency of robot deployment in various industries.

Possible Gemini 2.0 testing by Alphabet

TestingCatalog blog has uncovered a mysterious model called “2.0-Pro-Exp-0111” in a recent user interface update, which could be Gemini’s next major release. The timing aligns with Alphabet’s planned unveiling of a new AI agent called Project Jarvis next month. This development suggests that Google is making significant strides in AI capabilities, which could have implications for businesses using Google’s AI services or competing with them in various sectors.

Mistral releases open-source content moderation platform

Paris-based AI startup Mistral has released an open-source platform for content moderation. The platform uses AI to spot and flag harmful content across nine categories and 11 languages. This development could be particularly useful for businesses managing online communities, social media platforms, or any digital spaces where user-generated content is prevalent, helping to maintain safer and more appropriate online environments.

ByteDance releases X-Portrait2 platform

TikTok parent company ByteDance has released a platform called X-Portrait2 that can use clips from iconic films to turn still portraits into expressive animations. This technology could open up new possibilities for businesses in digital marketing, entertainment, and content creation, allowing for the production of more engaging and interactive visual content.

Google’s Vids AI video maker rolling out

Google’s Gemini-powered Vids app is rolling out to most Workplace subscribers. The app allows users to easily create tutorials, training videos, and other types of media using AI. This tool could significantly benefit businesses by streamlining the process of creating instructional and promotional video content, potentially reducing production costs and time.

Pixel phones to use AI for call replies

Google’s Pixel phones will soon be able to use AI for more sophisticated call replies, including the ability to confirm appointments on behalf of users. This feature could enhance productivity for business professionals, automating routine communication tasks and potentially improving customer service capabilities for companies using Pixel devices.

AI-created portrait of Alan Turing sells for over $1M

A portrait of Alan Turing painted by “robot artist” Ai-Da has sold for over $1 million at a Sotheby’s auction, becoming the world’s first AI-created portrait to sell at this price. This sale marks a significant milestone in the AI art market and could have implications for businesses in the art and creative industries, potentially opening up new avenues for AI-generated content and its monetization.

ProfilePrint unveils AI platform for ingredient analysis

Singapore-based startup ProfilePrint has unveiled an AI platform that can analyze the quality of different ingredients, including coffee beans, tea leaves, and milk. This technology could be particularly valuable for businesses in the food and beverage industry, potentially improving quality control processes and enhancing product consistency across supply chains.

AI Tools

🎨Looka: An AI-powered platform that creates custom logos and brand kits in minutes.

🖼️Flair: A design tool for creating product photos by dragging and dropping images into pre-made scenes.

📝Canva’s AI Sketch and Draw: Transforms rough sketches into polished images with multiple variations.

🎨OpenArt.ai: A free tool for generating AI images from rough sketches, allowing users to specify style and color scheme.

🎬Genbler: A photo and video AI SaaS solution for content creators, streamlining media creation.

🛠️AI App Generator: Build fully functional AI wrappers with backend API routes in seconds.

📊Diaflow: Provides powerful AI automation, apps, and internal workflow solutions for companies.

🔍FlowScraper: Automates websites and extracts data with no coding required.

🗣️Maibrain: Offers a way to preserve the voice and experiences of loved ones for future interaction.

🌐Surfshark: An industry-leading VPN for fast, secure online browsing when traveling.

📚SoBrief: A tool offering summaries of over 73,530 books in 40 languages using AI.

🤝TeamPal: Creates AI teams without any coding required, streamlining team assembly.

📖Meco: Delivers key takeaways from newsletters, providing a personalized audio briefing.

📄PaperGen: Generates well-structured long-form papers with citations and AI detection.

🤖FullContext: A unified AI-powered chatbot that engages, qualifies, and demos leads.

📬FyxerAI: A tool that organizes your inbox and drafts emails in your tone of voice.

🔧Zappit AI: Uses squads of AI agents trained to fix technical errors and generate content.

🖌️TARS AI Logos: Crafts the perfect Midjourney prompt for your brand’s logo using AI.

AI Jobs

🌱The Rundown – Head of Growth: Seeking a candidate to drive the company’s growth initiatives and expansion strategies.

☁️Lambda – Engineering Manager, Cloud: Oversees cloud infrastructure and engineering teams to ensure robust service delivery.

🔧Coreweave – Principal Engineer: Responsible for leading engineering projects and developing high-performance solutions.

📊Meta – Quality Data Analyst: In charge of analyzing data quality and ensuring accuracy across various datasets.

🖌️Grammarly – Principal Product Designer: Designs intuitive product interfaces and user experiences with a focus on creativity.