🗞 Dire Wolves Resurrected, Google Workspace AI Boost, and Nvidia's 253B Open-Source Model

🔎 The Latest on the AI Frontier:

• Colossal Biosciences Revives Dire Wolves After 12,500-Year Extinction

• Google Workspace Gets AI Upgrade with Automation Flows and Content Tools

• Nvidia Releases Open-Source 253B Parameter Model Outperforming Larger Competitors

• Meta Faces Antitrust Trial That Could Force Instagram and WhatsApp Divestiture

• Former Meta Exec to Testify Company's AI Model Boosted China's Capabilities

• Other news you might find interesting

🧬 Dallas-based Colossal Biosciences announces successful de-extinction of dire wolves after 12,500 years of extinction

• The company has birthed three dire wolf pups (Romulus, Remus, and Khaleesi) with a record-breaking 20 unique precision germline edits, including 15 edits from ancient gene variants that haven't existed for over 12,000 years.

• Scientists extracted and sequenced ancient DNA from a 13,000-year-old tooth and a 72,000-year-old skull, creating a genome with 500 times more coverage than previously available.

• In addition to dire wolves, Colossal has birthed two litters of cloned red wolves, the most critically endangered wolf species with fewer than 20 remaining in North America, using a new approach to non-invasive blood cloning.

🔧 Google enhances Workspace with new AI automation and content generation capabilities

• Introducing "Workspace Flows" to automate multi-step processes using plain language prompts, with planned third-party integrations beyond Google's ecosystem.

• Google Docs will soon convert drafts into podcast-style overviews and offer "Help me refine" suggestions to strengthen arguments and improve structure.

• Additional AI features include Google Sheets' "Help me analyze" for trend spotting, Google Meet's call summarization, and Google Vids' video clip generation using Google's Veo 2 model.

🖥️ Nvidia releases fully open-source Llama-3.1-Nemotron-Ultra-253B-v1 model that outperforms DeepSeek R1 at less than half the size

• The 253-billion parameter dense model features toggle ON/OFF reasoning capabilities, achieving 97% on MATH500 and 72.5% on AIME25 benchmarks with reasoning enabled.

• Designed for efficient inference, the model's architecture includes skipped attention layers and fused feedforward networks, allowing deployment on a single 8x H100 GPU node.

• Released under the Nvidia Open Model License, it supports input/output sequences up to 128,000 tokens and is compatible with multiple languages including English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai.

💼 Meta heads to trial next week as the FTC pursues antitrust case that could force divestiture of Instagram and WhatsApp

• Despite Mark Zuckerberg's repeated White House visits and attempts to persuade President Trump to settle, FTC Chair Andrew Ferguson stated: "I think that the President recognizes that we've got to enforce the laws."

• Central to the case is Meta's acquisitions of Instagram (2012) and WhatsApp (2014), with the FTC citing an email where Zuckerberg wrote about Instagram: "The potential impact is really scary and why we might want to consider paying a lot of money for this."

• The 8-week trial will feature testimony from Zuckerberg, former COO Sheryl Sandberg, and executives from rival companies, with Judge James Boasberg already cautioning that the FTC "faces hard questions about whether its claims can hold up in the crucible of trial."

🔍 Former Meta executive set to testify that Meta's Llama AI model helped accelerate China's AI capabilities, particularly contributing to DeepSeek's development.

• Sarah Wynn-Williams will tell the US Senate Judiciary Committee that Meta's "Project Aldrin" was a secret mission to enter China, including building "a physical pipeline connecting the United States and China" despite warnings it could provide backdoor access to Chinese authorities.

• DeepSeek, actively endorsed by Chinese officials, has emerged as a formidable rival to OpenAI at a fraction of the cost—launching for just $6 million compared to typical LLM development expenses.

• The testimony alleges Meta began briefing Chinese Communist Party officials as early as 2015 on critical emerging technologies, with an "explicit goal being to help China outcompete US companies" by "promoting the China Dream."

More news you might find interesting:

• Investigation claims over 100 former IDF soldiers and Israeli intelligence operatives work at Meta, including its head of AI policy.

• IETF forms AI Preferences Working Group to develop standards for expressing content usage permissions to AI scraper bots.

• AI startup founder reprimanded after attempting to use an artificially generated avatar to present arguments in a New York court hearing.

• Microsoft employees disrupt company's 50th anniversary celebration with protests over Israel military contracts.

• New Pew Research survey reveals US adults view AI more as a threat than a benefit, highlighting a unique opportunity for Apple.

• Verizon reports nearly 40% sales increase after implementing Google AI assistants for its customer service representatives.

Have any feedback? Send us an email