Today:

• OpenAI Joins Apple

• Google's $75M Fund Boosts AI Learning

• Oracle Enhances Cloud CX with AI

• Robots Learn Safe Falling Techniques

• Pope Advocates AI Rules at G7

• China's Vidu Rivals OpenAI's Sora

• DHS Forms AI Safety Advisory Board

BREAKING AI NEWS - OpenAI and iPhone | Sam Altman and Homeland Security | Elon Musk and Helen Toner.

Sam Altman, OpenAI's CEO, is experiencing a significant year with several developments. Recently, Reuters reported that Apple is negotiating with OpenAI to integrate its generative AI technology into the iPhone, potentially for the upcoming iOS 18. This move is part of Apple's shift towards AI technology after discontinuing its car project. 

Amidst these corporate shifts, Elon Musk and Mark Zuckerberg are notably absent from a new AI Safety and Security board established by the Department of Homeland Security, which includes representatives from major tech companies but deliberately excludes social media firms. Meanwhile, OpenAI faces internal challenges with employee concerns and leadership changes.

WATCH THE VIDEO ON YOUTUBE

Grow with Google launches new AI Essentials course to help everyone learn to use AI

Google has introduced the AI Essentials course through its Grow with Google initiative, aimed at equipping people from various backgrounds with AI skills to enhance their productivity, with no prior experience necessary. This self-paced course is supported by a $75 million AI Opportunity Fund from Google.org, which will benefit over 1 million Americans, including veterans and Goodwill Industries. 

Participants will learn directly from Google's AI experts and engage in practical applications like brainstorming, decision-making, and streamlining daily tasks. Upon completion, they receive a certificate. The initiative also collaborates with employers and educational institutions to integrate these skills into their programs.

GOOGLE

Oracle doubles down on generative AI with new features to accelerate enterprise deal cycles

Oracle is enhancing its Fusion Cloud Customer Experience (CX) with new generative AI features to help automate and speed up critical workflows in sales, marketing, and customer service. These updates, immediately available, build on Oracle's recent integration of generative AI across its cloud applications, marking a strong commitment to the technology. The AI tools will assist in creating responses for customer queries, optimizing schedules for service agents, and generating targeted content for marketing. 

Oracle emphasizes privacy and security, ensuring customer data remains protected within these AI applications. This move is part of Oracle's broader strategy to embed over 50 generative AI capabilities across its suite of applications, aiming to boost productivity, reduce costs, and create new revenue streams for businesses.

VENTUREBEAT

Humanoid robots are learning to fall well

Boston Dynamics and Agility Robotics are focusing on teaching their humanoid robots how to manage falls effectively. Learning to fall safely is crucial as robots, like humans, encounter unexpected situations that may lead to falls, especially in real-world applications. These robots are being trained not just to perform tasks but also to handle falls without sustaining significant damage, which is essential for their long-term functionality and reliability. 

This training includes techniques similar to human strategies, such as protecting crucial components and using the body to control the direction of the fall. 

TECHCRUNCH

Pope will attend G7 summit to discuss AI, Italy says

Pope Francis is set to attend this year's G7 summit in Italy, marking the first time a pontiff will participate in the summit's discussions. Scheduled from June 13-15 in Puglia, the summit will focus on significant global issues, including the challenges and ethical considerations surrounding Artificial Intelligence (AI). 

Pope Francis has expressed concerns about the potential dangers of AI, advocating for global regulations to ensure its use benefits the common good. 

REUTERS

Chinese team unveils first text-to-video AI model on par with Sora

Chinese tech firm ShengShu-AI and Tsinghua University introduced a new AI model called Vidu at the Zhongguancun Forum in Beijing on April 27, 2024. This model can create a 16-second video in high quality with just one click and is comparable to the US-based OpenAI’s model, Sora. 

Vidu uses a new technology called Universal Vision Transformer, which combines features of two different AI models to improve video creation. It is capable of generating detailed and dynamic video scenes that include unique Chinese cultural elements. 

GLOBAL TIMES

US Homeland Security names AI safety, security advisory board

The U.S. Department of Homeland Security has formed a new advisory board comprising leaders from top tech companies like OpenAI, Microsoft, and Google, along with other industry giants. Announced on April 26, the board is tasked with developing strategies to protect critical infrastructure, such as transportation and power grids, from AI-related threats. 

Secretary Alejandro Mayorkas emphasized the board's practical focus on implementing AI safely across essential services including energy, defense, and transportation. This initiative follows a DHS report warning of potential large-scale cyber attacks facilitated by AI, highlighting the urgency of securing critical national infrastructure.

REUTERS

🧠RESEARCH

Layer Skip: Enabling Early Exit Inference and Self-Speculative Decoding

The paper introduces LayerSkip, a method to make language model responses faster. It uses a special training method that lets some layers of the model be skipped during use, speeding up the process. This technique also includes a way to check and correct early responses, ensuring accuracy while using less memory.

How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suites

The paper discusses InternVL 1.5, an open-source multimodal language model designed to match the performance of commercial models. It includes improvements like a strong vision encoder, dynamic resolution adjustment for high-quality image inputs, and a bilingual dataset for better OCR and Chinese language tasks. The model performs comparably to both open-source and commercial models in tests.

Make Your LLM Fully Utilize the Context

The paper presents FILM-7B, a model designed to improve how large language models use long text inputs. It introduces a new training method that helps the model pay attention to important details throughout a long piece of text, which helps it understand and use the context better. This model has shown improved performance on tasks that involve understanding longer documents while maintaining its ability to handle shorter texts.

Tele-FLM Technical Report

The paper introduces Tele-FLM, a 52 billion-parameter multilingual language model optimized for efficient scaling and enhanced factual judgment. It performs comparably to larger models in multilingual tasks with fewer computational resources. The authors also share detailed methodologies, design insights, and training practices, aiming to support further academic and industrial advancements in developing large language models.

Interactive3D: Create What You Want by Interactive 3D Generation

The paper introduces Interactive3D, a new framework designed to enhance user control in 3D object generation. Unlike previous methods that rely heavily on 2D images or limited textual commands, Interactive3D allows users to interact directly with the 3D model. It provides tools for adding or removing components, manipulating shapes through dragging or geometric transformations, and making detailed semantic edits. This approach significantly improves the precision and quality of 3D creations, offering a more customizable and user-driven experience.

🛠️TOP TOOLS

Term Typer - Learn to type in a new language

Compass by Cohere - A New Multi-Aspect Embedding Model

Vetted - AI shopping agent for people who want the best quality products for their money – without endless research

Martin - A personal AI you can text, email, or talk to.

Texta - An AI article generator 

📲SOCIAL MEDIA

🗞️MORE NEWS

Yoneda Labs raises $4M from Khosla Ventures to build the ‘OpenAI for chemistry’

Yoneda Labs, supported by Y Combinator, raised $4 million to enhance drug discovery using AI. This funding, led by Khosla Ventures, will help automate chemical reactions to train AI models, speeding up and reducing costs in chemical manufacturing. The startup aims to develop a foundational AI model for chemistry, improving how new drugs are made. VENTUREBEAT

MongoDB CEO Dev Ittycheria talks AI hype and the database evolution as he crosses 10-year mark

Dev Ittycheria, CEO of MongoDB, has significantly evolved the company over his 10-year tenure, guiding it from a $30 million revenue base to nearly $2 billion. Under his leadership, MongoDB transitioned to cloud services, successfully launched an IPO, and expanded from a few hundred customers to nearly 50,000. Ittycheria also led MongoDB's shift from open source to a more controlled licensing model, launched a venture capital arm, and navigated a significant security breach.  TECHCRUNCH

AI for drug discovery draws a $1 billion launch—and a lot of hope

Xaira Therapeutics, a new AI-driven company aiming to revolutionize drug discovery, launched with a whopping $1 billion in funding, a record for its investor, Arch Venture Partners. With expertise from Genentech’s former chief scientific officer, the San Francisco-based startup plans to use AI to link biological targets with diseases, potentially speeding up the creation of life-saving drugs. FORTUNE

The Estée Lauder Companies and Microsoft increase collaboration to power prestige beauty with generative AI

The Estée Lauder Companies (ELC) and Microsoft have announced the creation of an AI Innovation Lab, marking an expansion of their strategic partnership. This lab will utilize Microsoft's Azure OpenAI Service to innovate in the beauty industry by enhancing consumer connections and accelerating product time-to-market with a focus on local relevance. The collaboration includes developing an AI-driven chatbot for internal marketing and leveraging AI in product development to quickly adapt to trends. MICROSOFT

‘To the Future’: Saudi Arabia Spends Big to Become an A.I. Superpower

Saudi Arabia is investing heavily in artificial intelligence (AI) as part of its Vision 2030, aiming to diversify from its oil-reliant economy and become a major player in global tech. The kingdom has created a $100 billion AI fund, and recently hosted the Leap tech conference in Riyadh, drawing over 200,000 participants, including top executives from Amazon, Google, and TikTok. Deals worth over $10 billion were signed at the conference, highlighting Saudi's strategic push into tech to compete with global powers like the US and China.

THE NEW YORK TIMES

AI just helped us find out where Plato is buried - here's how

Researchers at the University of Pisa used AI to decipher a charred papyrus scroll, revealing philosopher Plato's burial in Athens. The breakthrough is part of the Vesuvius Challenge, aiming to decode ancient scrolls buried by Mount Vesuvius's eruption, enhancing our understanding of history with AI technology. ZDNET