Today:

• Meta To Integrate AI To Ray-Ban Glasses

• Google.org $20M AI Accelerator Launch

• Microsoft Blocks AI Exploits

• Oracle Enhances NetSuite

• Amazon's $150B AI Investment

Meta is adding AI to its Ray-Ban smart glasses next month

Meta plans to integrate AI into its upcoming Ray-Ban smart glasses, set for release next month. The glasses boast features like object and animal identification, monument recognition, and text translation. Activating the smart assistant is as simple as saying "Hey Meta" followed by a prompt or query. 

While the AI showed promise in identifying pets and artwork, it stumbled with distant zoo animals and struggled with exotic fruits. Currently, the glasses support translations in English, Spanish, Italian, French, and German. Meta aims to refine these features further, available initially through a US-only early access waitlist.

THE VERGE

Google.org launches $20M generative AI accelerator program

Google.org is launching a $20 million program called Google.org Accelerator: Generative AI to support nonprofits developing technology using generative AI. Twenty-one nonprofits, including Quill.org and World Bank, will receive funding and access to technical training, workshops, mentors, and AI coaching. Teams of Google employees will work with selected nonprofits full-time for up to six months. 

Generative AI has the potential to increase productivity, creativity, and effectiveness in social impact efforts. Despite the benefits, many nonprofits face barriers such as lack of awareness, training, and funding, hindering adoption of AI solutions.

TECHCRUNCH

Microsoft Launches Measures to Keep Users From Tricking AI Chatbots

Microsoft introduced tools to safeguard AI chatbots from manipulation by malicious users. These tools, part of Azure AI, counter "prompt injection" attacks that aim to exploit AI systems for harmful purposes. Sarah Bird, Microsoft's Chief Product Officer, highlighted the importance of maintaining AI system integrity and user trust. 

The new offerings include prompt shields to detect and block attacks, and safety messages to guide responsible AI behavior. The move underscores the tech giant's commitment to AI ethics amid growing competition in the field. Collaboration, ethics, and computational power are pivotal in advancing AI technology.

PYMNTS

Oracle gives NetSuite an AI upgrade

Oracle is enhancing its NetSuite software with over 200 AI features, aiming to boost productivity across various business tasks without extra charges. The move challenges competitors who charge premiums for similar services and emphasizes practical AI applications for customer needs. AI integration into business systems is seen as inevitable, driving Oracle's commitment to accessibility. 

The focus is on simplifying tasks like financial data summarization and job description drafting. Oracle assures customer data privacy and anticipates future AI advancements, aiming for seamless user experiences. The initiative reflects a broader trend of AI adoption driving productivity gains in the economy.

AXIOS

Amazon Bets $150 Billion on Data Centers Required for AI Boom

Amazon is investing nearly $150 billion in data centers over the next 15 years to meet the growing demand for AI applications and digital services. The move is aimed at solidifying its position in the cloud services market, where it currently holds a significant lead. Amazon's spending spree includes expanding existing facilities and establishing new ones in locations such as Mississippi, Saudi Arabia, and Malaysia. 

The company's focus on AI-related projects underscores the importance of computing power in driving advancements in artificial intelligence. However, Amazon faces challenges such as securing sufficient electricity and addressing opposition to data center expansion. Despite these obstacles, Amazon remains committed to sustainable practices and renewable energy initiatives.

YAHOO!

🧠RESEARCH

ViTAR: Vision Transformer with Any Resolution

ViTAR is a solution addressing Vision Transformers' limited scalability across various image resolutions. ViTAR employs dynamic resolution adjustment and fuzzy positional encoding to maintain performance across resolutions. Results show adaptability with 83.3% accuracy at 1120x1120 and 80.4% at 4032x4032 resolutions, reducing computational costs while enabling versatile high-resolution image processing.

Mini-Gemini: Mining the Potential of Multi-modality Vision Language Models

Mini-Gemini is a framework aimed at enhancing multi-modality Vision Language Models (VLMs). Mini-Gemini addresses the performance gap compared to advanced models like GPT-4 and Gemini by focusing on high-resolution visual tokens, high-quality data, and VLM-guided generation. It enables improved image understanding, reasoning, and generation simultaneously, supporting various language model sizes and achieving leading performance in zero-shot benchmarks.

Long-form factuality in large language models

The paper presents a method, Search-Augmented Factuality Evaluator (SAFE), utilizing Large Language Models (LLMs) to assess long-form factuality. Using GPT-4, they created LongFact, a dataset for benchmarking factual accuracy. SAFE leverages LLMs to break down responses, assess facts via search queries, and proposes an F1 score extension. Results show LLMs achieve superhuman rating performance, offering cost-effective evaluation compared to human annotators.

ObjectDrop: Bootstrapping Counterfactuals for Photorealistic Object Removal and Insertion

ObjectDrop is a method for photorealistic object removal and insertion in images, addressing limitations of diffusion models in respecting physical laws. ObjectDrop employs a counterfactual dataset capturing scenes before and after object removal, fine-tuning diffusion models to remove objects and their effects. Bootstrap supervision further enhances performance, surpassing prior methods in realism and scene consistency.

BioMedLM: A 2.7B Parameter Language Model Trained On Biomedical Text

BioMedLM is a 2.7 billion parameter language model trained specifically on biomedical text from PubMed abstracts and full articles. Despite its smaller size compared to models like GPT-4 and Med-PaLM 2, BioMedLM achieves competitive performance on biomedical question-answering tasks and patient inquiries, showcasing potential for transparent, privacy-preserving, and cost-effective applications in biomedicine. The model is available on the Hugging Face Hub.

🛠️TOP TOOLS

RenderNet - Generate images, just the way you want. Face Mastery, Pose Perfection in seconds.

Kiosk - Leverage WhatsApp at scale, build meaningful relationships with clients and grow your business.

Lummi - The best free stock photos and royalty free images. Powered by robots everywhere.

Brevian - AI for business users 

Offdeal - AI broker to find, vet, and engage off‑market acquisition targets for you

📲SOCIAL MEDIA

🗞️MORE NEWS

Microsoft delights IT admins by removing Copilot from Windows Server 2025

Microsoft pleasantly surprises IT admins by axing Copilot from Windows Server 2025. Initially shocked by its addition to the typically feature-free server OS, admins rejoice as Microsoft swiftly reverses course in response to feedback.  Eyes remain peeled for any last-minute sneak-ins before the official release.  THE VERGE

AI21 Labs juices up gen AI transformers with Jamba

AI21 Labs introduces Jamba, a novel gen AI model blending Mamba and transformer architecture. While not displacing current models, Jamba excels in generative reasoning tasks, promising enhanced efficiency and context handling. Its hybrid design tackles limitations of traditional transformers, offering potential in enterprise applications. Jamba's development heralds a new frontier in generative AI. VENTUREBEAT

Gemini updated to automatically start Google Maps navigation

Google enhances Gemini phone assistant by enabling automatic initiation of Google Maps navigation upon requesting directions. The update streamlines the hands-free experience by eliminating the need to manually start navigation. Additionally, voice commands for setting reminders and calendar entries are now available with Workspace extension enabled. 9TO5GOOGLE

Biden administration unveils new rules for federal government's use of artificial intelligence

The Biden administration introduces new regulations for federal AI usage, aiming to safeguard rights, ensure transparency, and mitigate risks. Agencies must publish AI systems used, assess associated risks, appoint a chief AI officer, and implement safeguards against discrimination. The move follows input from diverse stakeholders and aligns with global AI initiatives. USA TODAY

Robovision raises $42M to accelerate AI-powered industrial automation

Robovision, an AI-powered computer vision platform, secures $42M investment for industrial automation. The Belgian-based company offers a no-code solution, upgrading machinery and production lines with Vision AI. With notable success in agriculture, it aims to expand globally, focusing on the US market and R&D for sustainability. TECH EU

French company Valeo to use more Google Cloud AI tools

French automotive supplier Valeo announces increased utilization of Google Cloud AI tools, emphasizing their significance in the car industry. This move integrates AI into software development, vehicle design, and customer service. Amidst lowered sales projections, Valeo aims to enhance productivity and relevance with generative AI, partnering closely with Google Cloud. REUTERS

Chip Startup Celestial AI Lands Massive $175M Series C

Celestial AI, an optical interconnectivity startup, secures a substantial $175 million Series C funding led by US Innovative Technology Fund. Their photonic fabric platform separates compute and memory, enhancing AI processing efficiency. With investments from major players like AMD Ventures and Samsung Catalyst, Celestial AI aims to revolutionize AI computing. CRUNCHBASE

AI Design Software Startup Raises $20 Million To Expand Beyond Cars And Shoes

Vizcom, an AI design software startup, secures $20 million funding led by Index Ventures to expand its industrial design software beyond automotive and footwear sectors. Co-founded by Jordan Taylor and Kaelan Richards, Vizcom's software accelerates design processes, attracting big-name customers like Ford and New Balance. With plans to enhance AI capabilities and 3D modeling, Vizcom aims to revolutionize product design workflows. FORBES

AI advertising start-up valued at $4bn after fundraising

Advertising start-up The Brandtech Group has raised $115mn in new funds, valuing the company at $4bn. The funds will be used to disrupt the traditional advertising industry by using artificial intelligence (AI) to create marketing campaigns. The group, founded by former Havas CEO David Jones, aims to make marketing services better, faster, and cheaper using technology, including AI and machine-generated content. The shift towards using generative AI threatens jobs in the industry, but it can dramatically reduce the cost and time taken to create advertising campaigns. The Brandtech Group will use the new funds to add scale, make investments and acquisitions, and expand its AI tools to its influencer businesses. FINANCIAL TIMES