🗞 OpenAI's Sora Demos: Impressive Human Realism in AI-Generated Videos 🎥

🔎 The Latest on the AI Frontier:

• OpenAI's Sora Demos: Impressive Human Realism in AI-Generated Videos 🎥

• OpenAI Fights to Keep Co-Founder's Files Secret in Copyright Case ⚖️

• AWS CEO Discusses Oracle Partnership and Enterprise AI Strategy 🤝

• Elon Musk's Colossus: New AI Supercomputer Sparks Debate 🖥️

• Anthropic's Focus: Developing Safe and Ethical AI Products 🔬

• Meta's Accidental Leak: Quest 3S Headset Image Revealed Early 🥽

• Apple Loses Tax Battle: Ordered to Pay $14.4B to Ireland 💸

• iPhone 16 Series Unveiled: New AI Features and Design Changes 📱

• Amazon Tests AI Voice Cloning for Audible Narrators 🎙️

• C3.ai Faces Uncertainty as Key Partnership Nears Expiration 📊

• AI Tool Helps NYC Diners Bypass Restaurant Waitlists 🍽️

• Hugging Face Launches LightEval: Open-Source AI Evaluation Tool 🔍

• CardiaTec Raises $6.5M for AI-Driven Heart Disease Treatments 🫀

• AI Voice Tech Aids Congresswoman with Rare Disease 🗣️

• Singapore Proposes Ban on Deepfakes in Elections 🗳️

🎬 OpenAI's latest Sora demos showcase stunning human realism, but questions remain about its overall capabilities. Link

• Two new showcase videos demonstrate Sora's ability to generate consistent characters with complex motions and costume changes.

• Some tech leaders view Sora as merely catching up to competitors, highlighting the rapid advancements in AI video generation.

• Limited access to Sora makes it difficult to fully evaluate its performance, including potential flaws and consistency issues.

🤖 OpenAI fights to keep co-founder's files confidential in copyright lawsuit. Link

• OpenAI is battling to prevent files from ex-co-founder Ilya Sutskever being included in discovery for a high-profile copyright lawsuit.

• The lawsuit, brought by prominent authors like George R.R. Martin, claims OpenAI violated copyright by using books to train AI models.

• The case highlights complex issues around using copyrighted materials to train AI and could set precedents for future AI copyright disputes.

💼 AWS CEO discusses Oracle partnership and AI strategy. Link

• AWS and Oracle launch strategic partnership with Oracle Database@AWS, aiming to offer unique value to customers.

• AWS focuses on building an enterprise AI platform allowing customers to utilize multiple large-language models flexibly.

• Garman emphasizes AWS's long-term approach to AI, prioritizing a comprehensive platform over rushing out chatbot technology.

🖥️ Elon Musk's new AI supercomputer Colossus draws mixed reactions. Link

• xAI unveiled Colossus, a supercomputer reportedly running on 100,000 Nvidia H100 GPUs, far surpassing competitors like Meta's 16,000 chip system.

• Some tech leaders, like LinkedIn co-founder Reid Hoffman, view it as merely "table stakes" to catch up in the competitive AI field.

• Questions remain about Colossus's actual capabilities, power usage, and how it aligns with Musk's efforts to develop Tesla's own Dojo chip to reduce reliance on Nvidia.

🧠 Anthropic is focused on developing safe and ethical AI, with a strong emphasis on thoughtful product development and safety evaluations before releases. Link

• Krieger joined Anthropic to work on AI products at scale while maintaining a startup-like environment focused on safety and ethics.

• The company is currently focused on enterprise AI products like Claude for Work, aiming to make AI assistants that can function more like coworkers.

• Anthropic is working on improving AI models' ability to handle longer-term tasks and maintain context over extended interactions.

🥽 Meta accidentally leaks Quest 3S headset image in PC app ahead of official reveal. Link

• The leaked image shows a VR headset design blending elements of Quest 2 and Quest 3, with six front-facing cameras in two clusters.

• Rumors suggest the Quest 3S could be priced as low as $300 and may not include controllers, relying instead on gesture controls.

• Meta is expected to officially unveil the new budget-friendly headset at its Connect event starting September 25th.

💰 Apple loses 10-year tax battle, must pay $14.4B in back taxes to Ireland Link

• Europe's top court ruled against Apple, upholding a 2016 European Commission order for the company to pay €13 billion ($14.4B) in back taxes to Ireland.

• The case stems from allegations that Apple received illegal tax benefits in Ireland over two decades.

• Apple plans to take a $10 billion one-time tax charge in its Q4 2024 fiscal quarter as a result of the ruling.

📱 Apple unveils iPhone 16 series with new AI features, camera improvements, and design changes. Link

• iPhone 16 and 16 Pro models feature a new dedicated camera button, improved AI capabilities powered by A18/A18 Pro chips, and larger displays (6.3" for Pro, 6.9" for Pro Max).

• Apple Watch Series 10 announced with a thinner design, larger display, and new health features like sleep apnea detection.

• AirPods 4 introduced with noise cancellation, improved design, and USB-C charging case; AirPods Pro 2 and Max also updated.

🎙️ Amazon introduces AI voice cloning for Audible narrators in limited beta test. Link

• Amazon is inviting select Audible narrators to create AI-generated voice clones to speed up audiobook production.

• Narrators will retain control over which projects use their AI voice and can edit pronunciation and pacing.

• Compensation will be on a "title-by-title basis" using a royalty share model, though specific rates were not disclosed.

📊 C3.ai faces potential turbulence as key partnership nears expiration. Link

• Baker Hughes partnership, accounting for over 1/3 of C3.ai's revenue, set to expire in April 2025.

• Recent departure of Dan Brennan, who oversaw the partnership, raises questions about its future.

• C3.ai's Q1 2025 results showed mixed performance, with subscription revenue falling short at $73.5 million.

🍽️ AI tool helps NYC diners skip long waitlists for hot restaurants. Link

• The Shortlist, launched by Jenn Han and Josh Jung, offers day-of table availability by scraping Google Maps and Resy data.

• Users can search for restaurants based on specific criteria like conversation-friendly, romantic, or even suitable for breakups.

• Since its Aug. 29 launch, the tool has gained thousands of subscribers and over 10,000 searches, with plans to expand beyond Manhattan and Brooklyn.

🔍 Hugging Face unveils LightEval, an open-source tool revolutionizing AI model evaluation. Link

• LightEval offers customizable evaluation pipelines that support multiple devices and can scale from small to large deployments, addressing the growing need for tailored AI assessment.

• The tool integrates with Hugging Face's existing ecosystem, including 120,000+ shared models, promoting transparency and collaboration in AI development.

• By democratizing AI evaluation, LightEval empowers businesses across industries to ensure their models meet specific ethical and operational standards before deployment.

🧬 Cambridge spinout CardiaTec raises $6.5M for AI-driven heart disease treatments. Link

• CardiaTec aims to tackle cardiovascular diseases, the world's leading cause of death with 17.9 million annual fatalities.

• The company is building the "largest human heart tissue-multi-omics dataset" through partnerships with 65 hospitals in the UK and US.

• CardiaTec differentiates itself by focusing on cardiovascular diseases, which only 3% of AI-first companies are targeting in drug discovery.

🗣️ AI voice tech helps Congresswoman with rare disease continue her work. Link

• Rep. Jennifer Wexton used AI to recreate her voice after being diagnosed with progressive supranuclear palsy, a rare brain disease.

• Wexton debuted her AI-generated voice on the House floor in July, marking a historic moment for assistive technology in Congress.

• The experience has given Wexton new perspectives on AI's potential benefits for those with disabilities, while also recognizing the need for responsible use and regulation.

🗳️ Singapore proposes legislation to ban deepfakes in elections, safeguarding electoral integrity. Link

• The bill aims to prohibit publishing digitally manipulated content that realistically depicts candidates saying or doing things they didn't, applying to both AI and non-AI techniques.

• Corrective directions can be issued to take down offending content, with non-compliance punishable by fines or imprisonment.

• A code of practice for social media providers to combat manipulated content on their platforms is also planned, addressing broader concerns about deepfakes beyond elections.

Enjoying the newsletter? Please consider supporting us:

Have any feedback? Send us an email