AI NEWS YOU MISSED ❗#27 INSEAD AI
Suggested by INSEAD-ers. From the 1,500+ person global INSEAD AI private business leader community, regularly sharing AI-related articles, vids, etc, of interest. (alum lead)
AI content shared in the INSEAD AI Community this week
Nvidia CEO Jensen Huang talks blowout quarter, AI, inferencing
- Nvidia’s Impressive Financial Performance: Nvidia significantly surpassed analysts’ expectations in its fiscal first quarter, with a remarkable 127% year-over-year increase in data center revenue and a positive sales forecast driven by strong AI spending momentum.
- Launch of Blackwell Chip: Nvidia announced the upcoming release of its next-generation chip, Blackwell, designed for trillion-parameter AI models and generative AI tasks. This chip is expected to generate significant revenue and is more versatile and powerful than its predecessor, Hopper.
- AI Factory Concept: Nvidia’s CEO, Jensen Wong, emphasized that the company builds comprehensive AI factories, integrating CPUs, GPUs, and sophisticated memory systems. These AI factories are then disaggregated for deployment in various data center architectures.
- Supply Constraints and Demand: Despite the strong demand for Hopper and Blackwell chips, Nvidia faces supply constraints expected to last until next year. The company is working to increase supply to meet the vast demand for its AI technology.
- Automotive Industry Impact: Nvidia is making significant strides in the automotive sector, with Tesla leading in self-driving technology. The use of AI to process video data for autonomous driving is becoming standard, highlighting the growing importance of AI in understanding and interacting with the physical world.
Nvidia Announces a 10-for-1 Stock Split. Here’s What Investors Need to Know
Nvidia announced a 10-for-1 stock split to take place in June, aiming to make its shares more accessible.
90-Second Recap: Satya Nadella’s Keynote at Microsoft Build
- Microsoft Copilot Platforms: Microsoft has developed three platforms, with the first being Microsoft Copilot, an everyday AI companion designed to provide knowledge and expertise at your fingertips. This stack includes the newly introduced Copilot+ PCs.
- Windows Copilot Runtime: The Windows Copilot Runtime is anticipated to be as transformative for AI as Win32 was for the graphical user interface, highlighting its potential significance in the AI landscape.
- Advanced AI Infrastructure: Microsoft is the first cloud provider to offer general availability of VMs based on AMD ND MI300X AI Accelerator and the public preview of Cobalt-based VMs. Additionally, GPT-4o is now generally available on Azure AI following its initial testing on Azure OpenAI Service.
- Real-Time Intelligence in Fabric: Microsoft is leading the small language model revolution and introducing Real-Time Intelligence within Fabric, extending the capabilities of Copilot beyond personal assistance to team assistance and infrastructure operations.
- Copilot Studio for Custom AI Agents: With Copilot Studio, anyone can build copilots with agent capabilities, empowering developers and builders to create tailored AI solutions. Microsoft emphasizes that their innovations are aimed at enhancing the work of developers and builders rather than celebrating technology for its own sake.
Microsoft Build 2024: Everything Revealed in 9 minutes
- Microsoft Copilot Platforms: Microsoft has built three platforms, with the first being Microsoft Copilot, an everyday AI companion that provides knowledge and expertise at your fingertips. The Copilot stack allows for the development of AI applications and solutions, including the newly introduced Copilot+ PCs, the fastest AI-first PCs ever built.
- Copilot Stack Partnerships with Nvidia and AMD: Microsoft has a deep partnership with Nvidia, spanning the entire Copilot stack and offering Azure confidential computing on GPUs. They will be among the first cloud providers to offer Nvidia’s Blackwell GPUs and are also delivering general availability of VMs based on AMD MI300X AI accelerator, providing the best price-performance on GPT-4 inference.
- Partnership with OpenAI and GPT-4o: Microsoft’s strategic partnership with OpenAI has led to the availability of GPT-4o on Azure AI. This multimodal model, trained on Azure, can handle text, audio, image, and video inputs and outputs, and is now generally available for testing and integration into applications.
- Copilot Studio and Custom Workflows: Copilot Studio allows users to build custom copilots with agent capabilities, tailored to specific business processes and workflows. It includes connectors for various third-party applications, enabling the integration of data from multiple sources to enhance AI functionalities.
- Team Copilot and AI for Empowerment: Microsoft is extending Copilot beyond personal assistance to team assistance with the introduction of Team Copilot, which can facilitate meetings, manage projects, and collaborate within Microsoft Teams and other platforms. The focus remains on empowering every person and organization, emphasizing the practical use of technology to create meaningful experiences.
Microsoft Surface Copilot + PC Event: Everything Revealed in 13 Minutes
- Distributed AI Vision: The speaker emphasizes a future where AI is distributed, leveraging both cloud and edge computing to reduce latency, ensure privacy, and create devices that can understand and interact with the world around them.
- Introduction of Copilot+ PCs: A new class of Windows PCs called Copilot+ PCs is introduced, designed to integrate distributed AI across edge and cloud. These PCs are described as the most powerful Windows PCs ever built, featuring advanced AI capabilities with Microsoft Copilot.
- Performance and Efficiency: Copilot+ PCs are powered by a new systems architecture with CPU, GPU, and NPU, capable of 40 trillion operations per second. They are 58% faster than the latest MacBook Air with M3 processor and offer exceptional battery life.
- Enhanced AI Capabilities: Microsoft Copilot has been updated for better integration with Windows, offering advanced text, voice, and vision capabilities through GPT-4.0. It can assist in various tasks, provide suggestions, and interact naturally with users.
- New Surface Devices: The first Copilot+ PCs, including new Surface Laptop and Surface Pro, are introduced. These devices feature improved performance, longer battery life, new AI experiences, advanced displays, and flexible design, with pre-orders starting immediately and availability from June 18th.
Microsoft debuts new Copilot+ PCs using OpenAI’s GPT-4o while taking shots at Apple
Microsoft introduced the new Copilot+ PCs, powered by OpenAI’s GPT-4o, emphasizing their advanced AI capabilities and performance advantages over similar Apple devices.
Nadella Says Microsoft Having ‘Complete Rethink’ on Windows for AI
Microsoft’s CEO described the new AI-powered Windows as a complete re-imagination from the silicon up, incorporating neural processing units to locally process AI models, enabling features like the photographic memory “Recall,” the integrated “Copilot” assistant, and third-party developer enhancements, transforming the overall hardware and software experience.
We are taking Copilot to the next level. 🚀- Mustafa Suleyman
Microsoft introduces Copilot+ PCs, a groundbreaking evolution in AI-powered personal computing, offering real-time, intuitive interactions with AI companions capable of seeing, hearing, and assisting in daily life, marking the most significant update to Windows in its forty-year history.
Microsoft’s Consumer CMO Yusuf Mehdi on Copilot, new PCs and AI future
Microsoft’s Yusuf Mehdi highlighted the launch of new AI-driven Copilot+ PCs, which offer advanced features like real-time translation and enhanced performance with Qualcomm chips, starting at $999, and are positioned to compete with Apple’s MacBook.
Scarlett Johansson has lawyered up after expressing shock and anger over OpenAI’s release of a ChatGPT voice that sounded eerily similar to hers, despite her declining Sam Altman’s offer to voice the chatbot.
Sam Altman Parody: The Scarlett Johansson Drama
In a parody statement, “Sam Altman” clarifies that Scarlett Johansson’s team approached them for collaboration, but after choosing a different direction and using a voice similar to hers for a demo, she complained despite knowing their interest in her voice; ultimately, they decided to remove the voice to avoid further drama, emphasizing that it was a result of innovation, not imitation.
Sam Altman’s Naughty Side — The Information
Sam Altman, CEO of OpenAI, faces controversy after actress Scarlett Johansson accused the company of using a voice resembling hers for ChatGPT without her consent, highlighting Altman’s risky approach of “asking forgiveness, not permission,” which may jeopardize his leadership of a major tech company.
Jan Leike Reflects on His Journey and Urges OpenAI to Prioritize AGI Safety
Superalignment co-lead Jan Leike announced his departure from OpenAI, emphasizing the need for a stronger focus on AGI safety and preparedness, expressing gratitude for his team, and calling for a cultural shift within the company to prioritize the serious implications of developing smarter-than-human AI.
OpenAI’s rapid release of its multimodal GPT-4o model ahead of Google I/O, with its “flirty” tone reminiscent of the AI in “Her,” has sparked debates on anthropomorphizing AI and led to significant internal dissent, including the departure of key members from its superalignment team over concerns about the company’s priorities and safety measures.
The Real Significance of Ilya Leaving OpenAI — INSEADer Robert Maciejko quoted
OpenAI Chief Scientist Ilya Sutskever’s departure has sparked renewed concerns about the company’s alignment and safety priorities, following months of speculation and the recent resignations of key team members.
OpenAI’s New Integrations and Advanced Data Analysis Toolkit Enhancements — INSEADer Justin Tan
OpenAI has introduced new features allowing users to connect Google Drive and Microsoft OneDrive directly to ChatGPT, along with significant enhancements to the Advanced Data Analysis toolkit, making data visualization more interactive and aesthetically pleasing, thereby advancing data analysis accessibility for various professionals.
🚨 Silicon Valley Needs a Moral Backbone 🚨 — INSEADer Robert Maciejko
AI developers must prioritize safety and trust over reckless innovation to ensure the success and ethical growth of the industry, as highlighted by the stark contrast in trust levels among AI CEOs in a recent survey.
Leading in the Era of AI / Fireside chat — INSEADer lead
- Event Recap: The INSEAD AI fireside chat, moderated by INSEADer Karen Ko Thuet, featured guest speakers Natalia Burina and INSEADer Kevin Pereira.
- Guest Speakers: Kevin Pereira, Managing Director at Blue Artificial Intelligence and AI consultant, and Natalia Burina, AI product leader and investor at Storm Ventures, shared their extensive experiences in AI across various industries and organizations.
- Advice for Business Leaders: Kevin emphasized using AI as a tool to solve specific business problems rather than a one-size-fits-all solution. He advised businesses to identify challenges that AI can address, integrate it effectively with human talent, and focus on education and proper utilization of AI tools.
- AI’s Impact on Jobs: The speakers discussed AI’s transformative effect on jobs, noting that while AI will automate repetitive tasks, it will also create new roles requiring human-AI collaboration. They stressed the importance of continuous learning and adapting to new AI tools to stay relevant in the evolving job market.
- Strategic Implementation of AI: Natalia highlighted the need for an experimental approach to implementing AI, starting with a clear understanding of data and identifying business opportunities. She emphasized the importance of high-quality human annotations and proprietary data in building successful AI products and recommended businesses adopt AI-first strategies to stay competitive.
To fully realize the productivity benefits of generative AI and other technologies, Europe and the United States must enhance human capital and accelerate technology adoption, which will involve significant occupational transitions and demand for new skills, particularly in high-skill professions like STEM and healthcare, while low-wage workers face the greatest challenges in adapting to these changes.
Don’t Get Caught Up In The GenAI Hype — INSEADer Hamilton Mann in Forbes
While generative AI shows promise for boosting productivity, businesses should remain cautious and focus on achieving clear return on investment, emphasizing effectiveness, efficiency, and process maturity rather than getting swept up in the hype.
Meta plans to keep the weights of its 400B model private, hoping the move will go unnoticed.
The future of foundation models is closed-source (Opinion)
The future of foundation models is likely to be closed-source, as centralizing forces of data and compute make it difficult for both open and closed-source AI to dominate long-term, with closed-source models expected to create more economic and consumer value over the next decade.
Amazon AI feature in product reviews
Having tried the AI in the Amazon store and found it helpful for inquiries, will this feature become standard on all company websites?
The current AI-to-AI conversation capabilities are already impressive and chilling, and they represent the worst performance these machines will ever have, suggesting even more astonishing advancements in the future.
Exploring the Multimodal Capabilities of ChatGPT-4o at Viva Technology
At Viva Technology, Romain Huet, Head of Developer Experience at OpenAI, demonstrated the powerful combination of ChatGPT-4o with Sora for creating videos, scripting, voice-overs, and more.
Artificial Intelligence and the Wisdom of the Heart: Towards a Fully Human Communication — Vatican
Pope Francis, in his message for the 58th World Day of Social Communications, emphasizes the importance of maintaining human wisdom and integrity in the age of artificial intelligence, advocating for ethical regulation and the responsible use of AI to ensure it serves humanity and fosters genuine communication.
What’s Brewing at Slack? Controversy Over AI Training Policy
Slack is under scrutiny for allegedly using customer data to train AI models, but Salesforce clarified that this was a misunderstanding, asserting that no customer data is used to develop large language models (LLMs), and their Privacy Principles have been updated accordingly.
Thread on what is in the California AI bill
It might be the most important legislation on AI given many of the frontier companies are based there
Open source proponents upset about AI law going thru California approval process
Sixteen global AI tech companies, including those from the US, China, and the Middle East, have agreed to new safety commitments to ensure responsible AI development, marking a historic agreement announced at the AI Seoul Summit.
Foundation Model Transparency Index — Stanford, MIT, Princeton
The Stanford Institute for Human-Centered Artificial Intelligence (HAI) has released an updated version of the Foundation Model Transparency Index, revealing that while transparency among 14 major AI companies has improved — raising the average score from 37 to 58 out of 100 — significant gaps remain, particularly concerning data usage and downstream impact disclosure.
AI Can Discover Corporate Policy Changes in Earnings Calls
Researchers have found that using ChatGPT to analyze quarterly earnings call transcripts can accurately uncover corporate investment policy changes that the market often misses, suggesting AI’s potential to enhance investment research by efficiently extracting and interpreting complex, unstated information from public corporate communications.
Mapping the Mind of a Large Language Model — Anthropic
Anthropic researchers have made a significant advance in understanding AI models by identifying how millions of concepts are represented within Claude Sonnet, a large language model, using dictionary learning techniques, which could lead to making AI models safer in the future.
Using ideas from game theory to improve the reliability of language models
MIT CSAIL researchers have developed a “consensus game” using game theory to enhance the reliability and accuracy of language models, enabling them to outperform larger models by harmonizing generative and discriminative querying to improve text comprehension and generation.
We’ll need universal basic income — AI ‘godfather’
Geoffrey Hinton, the “godfather of AI,” asserts that governments will need to implement universal basic income to address job displacement caused by AI, while also warning about AI’s potential to pose existential threats if not properly regulated.
In “Outgrowing Capitalism,” Marco Dondi argues for rethinking the creation and allocation of money to address economic insecurity and proposes a new socio-economic model called Monetism to provide a pragmatic roadmap for societal transformation.
100 articles on generative AI — McKinsey
Since generative AI’s emergence in late 2022, McKinsey has published over 100 articles highlighting its potential to add $2.6 to $4.4 trillion in annual value, revolutionizing business and society, with top insights including productivity impacts and strategies for scaling AI effectively.
Sampsa Samila’s talk at SAP Lagoas Park highlighted that AI can outperform elite students in ideation, AI usage by BCG consultants improved work quality, key questions about AI’s impact on commoditization and new possibilities, and drew parallels between historical skepticism of writing and current views on AI’s role in learning.
AI Spending Patterns : It’s Not What You Think — Tomasz Tunguz
Despite the overall growth in AI spending, with card transaction volume for AI software services jumping 293% year-over-year, AI vendor growth rates have declined by 78% annually, indicating a nuanced pattern where customer counts and account retention vary significantly among companies, and market dynamics are influenced by rapidly evolving technology, shifting buyer preferences, and changing pricing strategies.
AI detects sex-related differences in brain structure
A new study from NYU Langone Health reveals that AI programs analyzing MRI scans can accurately detect sex-related differences in brain structure, particularly in white matter, which may enhance understanding and treatment of neurological and psychiatric disorders.
How AI is impacting policy processes and outcomes in Africa
The rapid adoption of AI by African consumers, educational institutions, and governments underscores the urgent need for robust AI governance measures, integrating AI policy with broader initiatives on privacy, security, and data access, and fostering collaboration among governments, regional organizations, and industry to ensure responsible and equitable AI development across the continent.
Social Choice for AI Alignment: Dealing with Diverse Human Feedback
This paper discusses how the field of social choice can address the challenge of aggregating diverse human feedback into consistent data for fine-tuning AI models, drawing on insights from a recent workshop on Social Choice for AI Ethics and Safety.
The Platonic Representation Hypothesis
The Platonic Representation Hypothesis posits that representations in AI models are converging toward a shared statistical model of reality, similar to Plato’s concept of an ideal reality, as evidenced by the increasing alignment in how neural networks across different domains and data modalities represent data.
The mystery over why human brains have shrunk over time
The mystery of why human brains have shrunk by about 13% over the past 100,000 years remains unresolved, with theories ranging from more efficient brain organization due to symbolic thinking and language, to climate change, the development of complex societies, nutritional deficiencies from agriculture, and even self-domestication.
I placed Stockfish (white) against ChatGPT (black). Here’s how the game went.
ChatGPT is effective for solving certain problems, but struggles with tasks like playing chess against engines like Stockfish, where it often makes illegal moves.
Upcoming live events — INSEADer lead
June 6— INSEAD Prof. Theos Evgeniou — Open Source AI, Ecosystems & National Adaptation: A View from the UAE (ONLINE — FREE, RSVP required)
June 6 — Private INSEAD AI session — Building the Next Generation of AI Leaders with two Egon Zehnder colleagues — Burcu Bicakci (also an INSEAD MBA) and Christoph Wollersheim (Info in the private INSEAD groups)
We’re going deep on topics like:
- AI Trends and Challenges
- Preparing for AI Disruption
- Evolving Roles
- Identifying AI Leaders — Key Qualities, Background and Experience
- Industry- and Geo-Specific Skills
- Talent Acquisition Strategies
- Compensation and Incentives
- Cross-Functional Collaboration
- Upskilling Leaders
- Board’s Role in AI Governance
Join for an insider view of the world’s hottest job market. (INSEADer private event.)
Other Upcoming Live Events
Solving for happy in the age of AI — May 30
Former Google X CBO Mo Gawdat and NYU Stern professor Scott Galloway explore how AI impacts happiness, examining its influence on habits, gratitude, and coping with life’s challenges, in a free, 1-hour fireside chat.
More Videos of the Week
For a broad selection of AI videos, check out the INSEAD AI YouTube Channel (regularly updated)
From Chess to Career: Unlocking GenAI’s Training Potential w/ Henning Piezunka, INSEAD
- Webinar Summary: The recent webinar with INSEAD Associate Professor Henning Piezunka delved into the transformative role of AI as a training tool, drawing from his research in the realm of chess computers and beyond.
- Skill Development: Gain insights into how AI can accelerate skill development and provide a competitive edge, especially for individuals lacking human training partners.
- Democratizing Impact: Explore real-world examples showcasing the democratizing impact of AI in training, with scalable solutions applicable across various industries.
- Balanced Learning: Understand the limitations of solely relying on AI for training and learn how to strike a balance between AI-powered learning and real-world practice.
- Optimal Performance: Uncover the nuanced dynamics of skill enhancement in an AI-driven world and discover strategies to harness the strengths of both AI and human intelligence for optimal performance.
John Schulman (OpenAI Cofounder & new head of safety) — Reasoning, RLHF, & Plan for 2027 AGI
- Pre-training vs. Post-training: Pre-training involves training models to imitate internet content, while post-training targets specific behaviors, refining the model to act as a chat assistant and produce useful, human-preferred outputs.
- AGI Coordination and Safety: Schulman discusses the necessity of industry-wide coordination if AGI arrives sooner than expected. He emphasizes the importance of careful deployment, pausing training if needed, and setting up monitoring systems to ensure safety.
- Future Capabilities and Tasks: Schulman predicts that models will soon be capable of handling complex, long-term projects autonomously, such as managing entire coding projects. This will involve training models to recover from errors and efficiently handle long-horizon tasks.
- Sample Efficiency and Generalization: Schulman highlights the progress in making models more sample efficient and capable of generalizing from limited examples. He gives examples of models generalizing behaviors across languages and multimodal data.
- Human Oversight and AI Deployment: Schulman envisions a future where AI significantly augments human work without entirely replacing human oversight. He discusses the balance between integrating AI into more aspects of work while ensuring safety and alignment with human values.
Brave New Words — Greg Brockman & Sal Khan
- Book Announcement: Sal from Khan Academy announced his second book, “Brave New Words About the Future of AI in Education and Work,” which includes interviews with key figures like Greg Brockman from OpenAI.
- OpenAI’s Mission: Greg Brockman discussed the origins and mission of OpenAI, inspired by Alan Turing’s ideas, aiming to create human-level intelligence (AGI) that benefits humanity.
- AI Safety: Brockman emphasized the importance of AI safety and how OpenAI integrates safety measures while advancing AI technology.
- Educational Collaboration: The collaboration between Khan Academy and OpenAI aims to leverage AI for personalized education, introducing tools like Khanmigo for enhanced learning experiences.
- Optimism for AI’s Future: Both Sal and Greg are optimistic about AI expanding human capabilities and transforming education, despite the potential challenges and risks involved.
Google CEO Sundar Pichai says AI search will actually help the web
- Sundar Pichai on AI’s Impact: Google CEO Sundar Pichai discussed AI’s role in enhancing Google Search and the broader web, emphasizing the positive changes AI can bring to online experiences.
- Introduction of AI Previews in Search: Pichai introduced AI previews in Google Search, highlighting how these changes aim to improve user interaction with search results by providing more contextual information.
- Challenges and Concerns: The rollout of AI previews has raised concerns among web publishers about potential traffic declines, with some viewing it as a threat to their online presence and business models.
- AI and Content Creation: Pichai acknowledged the emotional and practical challenges that AI-generated content poses for creators, emphasizing the need for fair use and compensatory mechanisms for original content creators.
- Future Vision of the Web: Pichai envisions a richer, more interactive, and multimodal web experience driven by AI, while recognizing the need to balance innovation with maintaining the integrity and quality of online content.
- Collaboration: Psychological safety is essential for effective collaboration on AI-related topics, enabling open sharing of ideas and exploration of innovative solutions without fear of judgment.
- Continuous Learning: Boards need to continuously improve their knowledge of AI, requiring regular training sessions and bringing in external experts to stay informed about AI developments.
- Inclusive Tools: Tools can facilitate discussions and gather input on AI decisions, ensuring that all board members’ perspectives are considered equally.
- Strategic Integration: AI initiatives should be tied to company goals and incorporated into the board’s strategic planning and evaluation processes for more relevant and actionable outcomes.
- Innovation Opportunities: Despite its complexity, AI offers significant opportunities for innovation and growth, making it crucial for boards to foster an environment of psychological safety to effectively address AI-related challenges.
Past editions — AI NEWS YOU MISSED
SUBSCRIBE TO GET UPDATES IN YOUR MAILBOX
Follow Robert Maciejko on LinkedIn or X (Twitter)
