Musk's Gigafactory Ambitions, Deloitte's GenAI Survey, Nvidia's Rubin Reveal, and More ...

Welcome to AI Weekly Breakthroughs, a roundup of the news, technologies, and companies changing the way we work and live.

xAI to Build 'Gigafactory of Compute' by Fall 2025

Elon Musk's AI startup, xAI, plans to build a supercomputer dubbed the "Gigafactory of Compute" by fall 2025, using 100,000 Nvidia H100 GPUs to enhance its AI chatbot, Grok. This project aims to create a supercomputer at least four times larger than the biggest existing GPU clusters, aiming to outpace rivals like OpenAI and Google. Despite Nvidia's upcoming H200 and Blackwell GPUs, xAI has opted for the H100s, reflecting its ambitious scale and urgency. The project, potentially involving a partnership with Oracle, highlights Musk's commitment to advancing AI technology through significant hardware investments and strategic collaborations.

Deloitte Survey Gets Real on GenAI

As organizations move past the initial excitement surrounding Generative AI, they are now focusing on realizing its vast potential. Deloitte's second quarterly global survey on Generative AI in enterprises reveals a shift towards prioritizing value creation and achieving tangible results. This report is based on survey findings and in-depth interviews with senior executives across various industries. Organizations are scaling up their Generative AI efforts from experimentation to large-scale deployments to maximize business impact and workforce integration. This transition involves overcoming significant challenges, including building trust in AI systems and evolving the workforce to adapt to new skills and roles. The report delves into these critical areas—value, scaling, trust, and workforce—to guide organizations in their Generative AI journey, with future surveys addressing additional challenges in AI scaling and value creation.

Surveys Asks: Are CXOs Ready for GenAI?

In the Wavestone Global Technology & Data Leaders Survey 2024, they talked to nearly 600 technology leaders – Chief Information Officers/Chief Technology Officers/Chief Digital Officers/Chief Information Security Officers – in Europe, North America, and Asia about the impacts of this seismic technology on all parts of their business – covering topics from sustainability to cybersecurity and from foundational capabilities to functional impacts. The findings reveal that many have yet to lay the necessary foundations to embrace the opportunities of GenAI – while avoiding the risks.

Nvidia's Jensen Huang Discloses Next-Gen Rubin Platform

Nvidia CEO Jensen Huang, in a keynote at National Taiwan University ahead of Computex, unveiled the next-generation Rubin platform and highlighted Nvidia's role in the AI-driven Industrial Revolution. He emphasized the integration of AI into industries, predicting massive economic impact and showcasing the new Blackwell Ultra and Rubin Ultraq GPUs using TSMC processes. Huang also discussed Nvidia's advancements in generative AI inferencing and the concept of "physical AI," where robots learn and interact autonomously in simulated environments. He praised Taiwan's crucial role in Nvidia's success and drew significant attention by engaging with local tech leaders and the community.

Apple Plans AI-Based Siri Overhaul to Control Individual App Functions

Apple plans to significantly enhance Siri with advanced AI in iOS 18, allowing users to control individual app functions via voice. Future updates will include sophisticated features such as chaining multiple commands, aimed at improving user interaction and efficiency.

OpenAI Signs Content Deals with The Atlantic and Vox Media

Sam Altman-led OpenAI said on Wednesday it has signed content and product partnerships with The Atlantic and Vox Media, helping the artificial intelligence firm to boost and train its products.

Mistral Introduces AI Non-Production License

Mistral AI introduces a new Non-Production License to balance openness and business Growth.

OpenAI Board Forms Safety and Security Committee

OpenAI has established a new Safety and Security Committee, led by Bret Taylor (Chair), Adam D'Angelo, Nicole Seligman, and CEO Sam Altman. This committee will focus on making critical safety and security recommendations for all OpenAI projects. The formation comes as OpenAI embarks on training its next frontier model, aimed at advancing capabilities towards achieving Artificial General Intelligence (AGI). The committee's immediate task over the next 90 days is to evaluate and enhance OpenAI’s safety processes and safeguards. Following this period, their findings and recommendations will be reviewed by the full board and then shared publicly. This initiative underscores OpenAI's commitment to leading in both capabilities and safety in the AI industry.

Microsoft Announces a Copilot for Telegram

Copilot, powered by GPT, now integrates seamlessly with Telegram, offering users a smarter chat experience with AI-driven assistance. From gaming strategies and entertainment recommendations to culinary guides and personalized playlists, Copilot caters to diverse interests, enhancing daily interactions and information access within the Telegram app. Whether planning a trip, seeking fitness advice, or exploring new music, Copilot aims to enrich user engagement by providing tailored responses across various categories.

AMD Announces New AI chips Amid Intensifying Competition with Nvidia, Intel

At Computex in Taipei, AMD unveiled new AI chips as part of its aggressive strategy to compete with Nvidia and Intel. CEO Lisa Su emphasized AI as the company's top priority, announcing the Ryzen AI 300 series for AI laptops, and the Ryzen 9000 series for desktops, both launching in July. AMD also revealed its Instinct MI325X accelerators for data centers, set for release in Q4, with the MI350 series in 2025 and MI400 series in 2026. Additionally, Su previewed the fifth-generation EPYC server processors due in the second half of the year. These chips, built on the new Zen 5 architecture, aim to enhance performance across supercomputers, data centers, and PCs.

Introducing Perplexity Pages

Meet Perplexity Pages, a new tool for easily transforming research into visually stunning, comprehensive content. Pages streamlines the process of crafting in-depth articles, detailed reports, or informative guides.

Codestral: Democratizing coding with Mistral AI

Codestral, Mistral's first-ever code model, is an open-weight generative AI model explicitly designed for code generation tasks. It helps developers write and interact with code through a shared instruction and completion API endpoint. As it masters code and English, it can be used to design advanced AI applications for software developers.

Claude Can Now Use Tools

Tool Use, which enables Claude to interact with external tools and APIs, is now generally available across the entire Claude 3 model family on the Anthropic Messages API, Amazon Bedrock, and Google Cloud's Vertex AI. With tool use, Claude can perform tasks, manipulate data, and provide more dynamic—and accurate—responses.

MavenAGI Launches AI Customer Support Agents Powered by OpenAI

MavenAGI recently launched an AI customer service agent, built on the flexibility of GPT-4, which a number of companies like Tripadvisor, Clickup and Rho are already using to save time and better serve their customers.

Scale Launches Expert-Evaluated LLM Leaderboards

Scale's SEAL Research Lab introduces expert-driven, reliable LLM leaderboards with private datasets to ensure fair model comparisons, focusing on coding, instruction following, math, and multilinguality, aiming to boost AI transparency and development.

What We Learned from a Year of Building with LLMs (Part II)

Key insights from a year working with LLMs, exploring operational tactics and long-term strategies for successful applications in the tech sector.

How A.I. Made Mark Zuckerberg Popular Again in Silicon Valley

After some trying years during which Mr. Zuckerberg could do little right, many developers and technologists have embraced the Meta chief as their champion of “open-source” artificial intelligence.

SignLLM: Sign Languages Production Large Language Models

This paper introduces the first comprehensive multilingual sign language dataset named Prompt2Sign, which builds from public data including American Sign Language (ASL) and seven others. The dataset transforms a vast array of videos into a streamlined, model-friendly format, optimized for training with translation models like seq2seq and text2text. Building on this new dataset, the authors propose SignLLM, the first multilingual Sign Language Production (SLP) model, which includes two novel multilingual SLP modes that allow for the generation of sign language gestures from input text or prompt. Both of the modes can use a new loss and a module based on reinforcement learning, which accelerates the training by enhancing the model's capability to autonomously sample high-quality data. The authors present benchmark results of SignLLM, which demonstrate that our model achieves state-of-the-art performance on SLP tasks across eight sign languages.

EthonAI Raises $16.5M

AI manufacturing startup funding is on a tear as Switzerland’s EthonAI raises $16.5M

Google Plans $2 Billion Malaysia Data Center

Alphabet’s Google will invest $2 billion to establish its first data center in Malaysia to power new cloud services, the latest in a string of multibillion-dollar plans by Western tech giants to meet growing computing needs in Southeast Asia.

China’s $47B Semiconductor Fund Puts Chip Sovereignty Front and Center

China has initiated a massive $47 billion state-backed investment fund, known as 'Big Fund III', to bolster its semiconductor industry and achieve greater chip sovereignty. This initiative aims to reduce reliance on foreign technology by enhancing both advanced and legacy chip production capabilities. This strategic move highlights China's intent to navigate the ongoing global tech tensions and establish a more self-sufficient semiconductor industry amidst competitive pressures from the U.S. and Europe.

Data Cloud Summit 24 - San Francisco - June 03 - 06

AI & Big Data Expo - California - June 05 - 06

Apple’s Worldwide Developers Conference - Cupertino - June 10 - 14

AI Engineer Summit - San Francisco - June 25 - 27

World Summit AI - Amsterdam - October 9 - 10

Gitex Global - Dubai - October 14 - 18

Big Data Conference Europe - Vilnius - November 19 - 22