🗞 OpenAI Partners with TSMC for Custom AI Video Chip 🎞️

🔎 The Latest on the AI Frontier:

• OpenAI Partners with TSMC for Custom AI Video Chip 🎞️

• Elon Musk's xAI Unveils Massive AI Training System 🚀

• Google's Gemini AI May Come to Android Auto 🚗

• Windows 11's Recall AI Feature Not Removable, Says Microsoft 🖥️

• DeepMind's GenRM: AI That Verifies Its Own Work 🧠

• Meta's New VR Headset Hinted in FCC Documents 🥽

• Gartner Predicts Hurdles for Generative AI Projects 📉

• AI-Generated Student Work Poses Detection Challenges 🎓

• Canva Raises Prices, Cites AI Features as Justification 🎨

• AI Model Predicts NFL Week 1 Outcomes for Bettors 🏈

• Studies Show AI May Enhance Rather Than Replace Workers 💼

• Lloyds Bank Partners with Cleareye.ai for Trade Finance 📊

• Clearview AI Faces Hefty GDPR Fine in the Netherlands 🔍

• Huawei Plans Product Launch to Rival iPhone 16 📱

🖥️ OpenAI reportedly developing first in-house chip with TSMC for AI video generation. Link

• Chip to use TSMC's advanced A16 Angstrom process, not expected in production until 2026.

• Custom chip aims to boost capabilities of OpenAI's Sora video generation AI.

• Move could potentially benefit Apple device sales through integration of advanced AI video features.

🤖 Elon Musk's xAI launches Colossus, now the most powerful AI training system. Link

• Colossus uses a 100,000 Nvidia H100 GPU cluster, surpassing other major AI models in scale.

• The system will double to 200,000 GPUs in the coming months, further increasing its capabilities.

• Developed in collaboration with Nvidia, Colossus is said to offer exceptional energy efficiency gains.

🚗 Google's Gemini AI may be integrating with Android Auto, code hints suggest. Link

• Code in latest Android Auto update includes Gemini-related elements like the sparkle icon and "GeminiLiveAssistantAction".

• Integration could allow users to interact with Gemini for intelligent responses while driving.

• Potential uses may include getting directions, finding rest stops, and receiving traffic updates.

🖥️ Microsoft confirms Windows 11 Recall AI cannot be uninstalled, option was a bug. Link

• Recent Windows update mistakenly showed option to uninstall Recall AI, which Microsoft says was a glitch.

• Users will be able to disable Recall AI once released, but not completely remove it from Windows 11.

• Recall AI rollout to Windows Insiders with Copilot+ still planned for next month.

🧠 DeepMind's GenRM improves LLM accuracy by having models verify their own outputs. Link

• GenRM trains verifiers using next-token prediction, leveraging LLMs' text generation capabilities.

• It outperforms standard approaches like discriminative reward models and LLM-as-a-Judge across various reasoning tasks.

• GenRM can generate chain-of-thought reasoning before verification, identifying subtle errors other methods miss.

• On the GSM8K math benchmark, a Gemma-9B model trained with GenRM solved 92.8% of problems, surpassing GPT-4 and Gemini 1.5 Pro.

🥽 FCC documents hint at imminent launch of new Meta VR headset, likely Quest 3S. Link

• FCC approval documents for a new Meta device were discovered, suggesting a launch may be near.

• The device, codenamed 2AGOZ-P97, supports Wi-Fi 6E like the Quest 3.

• Quest 3S is expected to be a cheaper version of Quest 3, replacing Quest 2 at around $300.

🚪 Gartner predicts one-third of generative AI projects will be abandoned by 2025. Link

• High upfront costs ($5-20 million) and difficulty measuring ROI are major challenges for companies deploying generative AI.

• Despite challenges, some companies report benefits like revenue increases, cost savings, and productivity gains from generative AI.

• Factors that could doom AI projects include inadequate risk controls and poor data quality, in addition to high costs.

🎓 Teachers still struggle to reliably detect AI-generated student work, as text checkers remain flawed. Link

• Commercial AI text detection tools have significant limitations, often failing to flag AI-generated text while incorrectly flagging human-written content.

• Experts suggest focusing on disclosure of AI use rather than detection, as machine-generated text becomes increasingly indistinguishable from human writing.

• The situation presents a challenge for educators to evolve teaching methods, similar to adaptations made for spell check, the internet, and Wikipedia.

📊 Canva justifies 300% price hike for Teams subscriptions with AI-powered features. Link

• US users face increase from $120 to $500 per year for up to 5 users, with 40% first-year discount.

• Australian teams could see annual costs rise from $480 AUS to $2,430 AUS for 5 users.

• Canva cites "expanded product experience" including Magic Studio and Visual Suite as reasons for the increase.

🏈 AI-powered model predicts NFL Week 1 game outcomes, offering betting insights. Link

• SportsLine's AI PickBot analyzes all 16 Week 1 games, providing against-the-spread, over-under, and money-line predictions.

• The AI model predicts the Broncos will cover as 6-point underdogs against the Seahawks, projecting a close 21.86-20.34 score.

• Last season, the AI PickBot successfully hit 1,674 4.5- and 5-star prop picks, demonstrating its accuracy in NFL predictions.

💪 AI may boost worker efficiency rather than replace jobs, according to recent studies and company experiences. Link

• Alorica introduced an AI translation tool allowing reps to handle calls in 200 languages, but is still actively hiring staff.

• IKEA retrained 8,500 customer service workers for more complex tasks after introducing AI chatbots for simple inquiries.

• A study found customer support agents using AI tools were 14% more productive, with the biggest gains (34%) for less experienced workers.

🤝 Lloyds Bank and AI platform Cleareye.ai launch trade finance partnership to streamline processing and compliance checks. Link

• The collaboration will use Cleareye.ai's ClearTrade technology to automate document examination and compliance checks for trade finance documentation.

• AI tools like optical character recognition, machine learning, and natural language processing will extract critical information from digital and paper-based trade documents.

• The partnership aims to help Lloyds' clients "trade simpler, faster and more efficiently" by streamlining key parts of trade finance processes.

🔍 Clearview AI hit with €30.5 million GDPR fine by Dutch regulator for privacy violations. Link

• The fine is Clearview's largest in Europe to date for its facial recognition database built without consent.

• Violations include collecting biometric data without a legal basis and failing to comply with data access requests.

• Dutch regulators are exploring holding Clearview executives personally liable to enforce compliance, as the company has ignored previous EU fines.

📱 Huawei plans product launch hours after Apple's expected iPhone 16 unveiling on September 10th. Link

• Huawei event scheduled for 2:30 PM Beijing time on September 10th, shortly after Apple's event.

• Huawei executive Richard Yu called it an "epoch-making product" that others "cannot make."

• The launch comes as Huawei challenges Apple in China, with Apple recently losing its top 5 smartphone vendor position there.

Have any feedback? Send us an email