AI NEWS YOU MISSED ❗#47

Top 5 stories of the week

1️⃣ Are Scaling Laws Fading in AI?

2️⃣ OpenAI Focus on Advanced Reasoning

3️⃣ Musk’s Supercomputer Jolts AI Rivals

4️⃣ OpenAI readies AI agent release

5️⃣ Anthropic — AGI by 2026/2027?

Top 5 stories of the week in detail

1️⃣ Are Scaling Laws Fading in AI?

Recent reports suggest OpenAI and rivals may be experiencing diminishing returns from scaling up their models, potentially challenging established scaling laws — which predict consistent performance gains with increased compute — and raising questions about the value of the billions invested in larger AI infrastructures.

Business Insider take:

• OpenAI’s upcoming model, Orion, shows only modest improvement over GPT-4, sparking industry concerns about an AI performance plateau.
• Experts debate whether scaling laws — traditionally guiding AI improvement through data and computing power — are hitting diminishing returns.
• Limited data availability and computing power constraints pose challenges to further advancing AI models.
• Some believe AI models may need new approaches beyond scaling to maintain innovation, as seen with OpenAI’s recent focus on inference improvements.
• The AI industry faces potential recalibration if future models struggle to achieve past performance leaps, impacting investment and growth expectations.

OpenAI’s Steve Heidl view — Integrating Today’s Tech Alone Will Take 10 Years

2️⃣ OpenAI Shifts Focus to Advanced Reasoning

READ (Paywall) 

Reuters article (Free)

OpenAI is shifting its focus from traditional GPT advancements to a new reasoning model, “o1,” as improvements in generative AI seem to be plateauing, sparking industry-wide discussion on whether AI’s progress is reaching a scaling limit and what this shift means for the future of AI development and infrastructure.

3️⃣ Musk’s Supercomputer Jolts AI Rivals

READ (Paywall)

Elon Musk’s rapid creation of xAI’s massive supercomputer, Colossus, has unsettled competitors like OpenAI, Microsoft, and others, igniting a high-stakes race to expand data center capacity and accelerate AI development.

4️⃣ OpenAI readies AI agent release

READ

OpenAI is expected to launch an AI agent named “Operator” in January, designed to autonomously execute tasks like coding and booking travel, marking a shift towards practical automation and aligning with an industry trend towards AI agents that enhance productivity by handling complex tasks.

5️⃣ Anthropic — AGI by 2026/2027?

Dario Amodei: Anthropic CEO on Claude, AGI & the Future of AI & Humanity | Lex Fridman Podcast (5:15 Interview)

1. Scaling and Progress: Amodei discussed the rapid advancement of AI scaling laws, predicting that AI capabilities could reach human-level cognition by 2026 or 2027, driven by scaling larger models, longer training, and vast data sets.
2. Anthropic’s Mission and Model Development: Anthropic’s approach centers on “race to the top” ethics, focusing on AI safety and interpretability. They develop models (Claude, Sonnet, Opus) across varying power levels to meet both complex and everyday tasks.
3. AI Risks and Responsible Scaling Policy: Amodei emphasized significant risks in AI, particularly catastrophic misuse (cyber, biosecurity) and autonomy risks, advocating for stringent safety and control measures as AI systems advance.
4. Limits of Large Language Models (LLMs): Amodei highlighted potential ceilings in scaling due to limitations in compute power, data quality, and diminishing returns on model accuracy, suggesting synthetic data as a potential solution.
5. Mechanistic Interpretability: Chris Olah’s work on understanding model mechanics was noted as a method to reverse-engineer AI behavior for safety, helping to ensure AI behaves predictably and transparently.

Happy Birthday to the INSEAD AI community

November 13 is the 1st birthday 🎂 of the INSEAD AI community. It’s incredible to see how far we’ve come over the past year.

Our community has grown into a valuable network for exchanging ideas, building knowledge, and keeping a pulse on the ever-evolving world of AI across various fields.

Thank you to everyone who has contributed to our discussions, shared insights, and supported one another’s learning journey. A special shout-out goes to our most active contributors, you’ve made our community even more valuable. We appreciate it.

Here’s to another year of advancing our understanding and application of AI together.

With appreciation and excitement for what’s to come, Robert and Fabio (prepared by humans)

ChatGPT Prompt of the week

“Based on what you know about me, draw a picture of what you think my current life looks like”

Results from INSEADers

You need to have Memory “on”

If you do not want your data to train models, you can shut that off here:

Then try “Based on this picture, describe the life of this person.”

Upcoming live events

INSEAD — BCG x — The Tipping Point of Technology — Nov. 19, 2024

Join INSEAD’s Annet Aris and BCG X’s Hanno Stegmann on November 19 to explore the emerging technology landscape and the unpredictable tipping points that shape society, industries, and companies, with insights into the factors influencing technology adoption in both B2C and B2B contexts.

Microsoft Ignite — live from Chicago — Nov. 19–22, 2024

CEO Satya Nadella and the Microsoft team will present various solutions, with a significant focus on AI advancements. Many sessions are free and will be recorded for later viewing.

INSEAD TECH TALK X — Blockchain Meets AI: Building Trust & Accountability in the Digital Age — Nov. 20, 2024

Join INSEAD Professor Peter Zemsky and industry experts on November 20 to explore the role of blockchain and AI in enhancing transparency, accountability, and ethical innovation, featuring real-world applications and future insights from leaders in the field.

Designing Human-Centric Organizations in the Algorithmic Age — Dec. 12, 2024

Join INSEAD Professor Phanish Puranam on December 12 to explore strategies for designing human-centric organizations in the age of AI, focusing on balancing technological efficiency with employee well-being, motivation, and meaningful engagement.

More AI stories of the week

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-led)

Where’s the Value in AI? — BCG Report

Only 4% of companies achieve substantial value from AI, according to BCG, as leaders focus on high-impact initiatives, strategic investments, and integrating AI across core processes, setting a roadmap for others to maximize their AI potential.

Golden Age of Building — Y-Combinator

With AI transforming work at an unprecedented pace, we’re entering a golden age of building, presenting unique opportunities to innovate in areas like government software, public safety, stablecoins, chip design, and more, aiming to create impactful solutions that strengthen the country.

Senior Managers Need to ‘Step Up’ on AI

Senior leaders must take a proactive role in understanding and guiding AI strategy, rather than relying solely on tech-savvy juniors, as AI risks and challenges demand strategic oversight beyond project-level solutions.

Salesforce hiring 1,000 AI sellers

Salesforce is hiring over 1,000 salespeople to drive growth for Agentforce, its new AI-powered customer service and sales tool, aiming to outpace rivals like Microsoft and ServiceNow in the autonomous AI agent market.

There’s An AI For That

AI tool for AI tools

IMF — AI Preparedness Index

The AI Preparedness Index (AIPI), developed by the International Monetary Fund (IMF), assesses 174 countries’ readiness for AI adoption across four key areas: Digital Infrastructure Human Capital, Innovation, Regulation 

The Global AI Index 2024: Measuring AI Progress Worldwide

The Global AI Index, now in its fifth edition, ranks 83 countries on their AI implementation, innovation, and investment levels, capturing the rapid advancements in global AI, including groundbreaking policies, technological races, and soaring investments that continue to drive AI’s transformative potential worldwide.

Gen AI: What Businesses Are Using It For

Businesses are increasingly using Generative AI to streamline tasks across Marketing, HR, and Operations, moving from experimental stages to targeted investments focused on ROI and freeing up time for more human-centric work.

Why Europe Risks Falling Behind in AI — and How to Catch Up

Europe risks falling behind in AI due to lower investment, infrastructure limitations, and high operating costs, potentially missing out on AI-driven economic growth unless it closes these gaps and boosts sector-specific capabilities.

AI Excels at Analyzing Harvard Business Cases — Ethan Mollick

Testing brand new Harvard Business School cases in various fields, Claude AI impressively handles most questions accurately — even surpassing case details in some instances — highlighting its potential in tackling complex MBA-level analysis.

Diablo Canyon Pioneers Generative AI for Nuclear Operations

Diablo Canyon Power Plant will become the first U.S. nuclear plant to implement an on-site generative AI solution, developed by Atomic Canyon with NVIDIA’s support, aimed at enhancing operational efficiency, reducing costs, and improving safety in nuclear energy management.

LLM Prompt Tuning Playbook

The “LLM Prompt Tuning Playbook” by Varun Godbole and Ellie Pavlick provides a comprehensive guide on effectively prompting post-trained large language models (LLMs) through a blend of theoretical frameworks and practical strategies, emphasizing the nuances of pre-training versus post-training and offering iterative methods to improve prompt clarity and relevance for enhanced LLM performance.

OpenAI opens French Office

Anthropic has hired an ‘AI welfare’ researcher

Anthropic has hired Kyle Fish as its first dedicated AI welfare researcher to explore moral considerations for AI, reflecting growing industry interest in whether AI systems might soon require ethical protections.

Famous artworks come to life with AI

WATCH ON YOUTUBE

Andreia Dominques from Web Summit 2024

With vibrant pink and purple colors setting the scene, preparations for Web Summit 2024 are well underway, with teams working tirelessly since late October to ensure a memorable opening.

The AI Revolution in Medicine: GPT-4 and Beyond

“The AI Revolution in Medicine” explores GPT-4’s transformative potential in healthcare, from enhancing diagnostics to supporting patient care, while urging careful management to balance AI’s profound benefits with its risks, making it essential reading for anyone impacted by the future of medicine.

Local School is teaching AI Business Development

Public Sector AI Week in Review (11/9/24) — Federal Innovator Chris Kraft 

AI Learning Resources

21 Days of Free Access to 365 Data Science’s Data & AI Courses

365 Data Science is offering free access to its full range of Data and AI courses for 21 days, including certificates, with no credit card required. This is a great chance for those interested in data or AI to explore top courses and deepen their expertise in these fields.

Fun

Last week in edition #46 we covered on initial views of the effect of US Presidential elections on AI. Today, 2 lighter takes.

An AI Generated alternative reality (Fun)

WATCH ON YOUTUBE

Elon Musk’s First Day at D.O.G.E. (Fun-ish)

WATCH ON X

Jobs

New AI Governance Jobs from Oracle’s Kevin Fumai

AI Videos of the Week

Top Minds in AI Explain What’s Coming After GPT-4o 

with Dr. Kai-Fu Lee, Chairman & CEO, Sinovation Ventures, CEO, 01.AI

Richard Socher, CEO & Founder, you.com, Co-Founder & Managing Director, AIX Ventures

Prem Akkaraju, CEO, Stability AI

1. Content Explosion Through Multimodal AI: In the next decade, AI will produce up to 20 times more content, especially 2-minute pieces, revolutionizing content creation across text, video, sound, and images, and increasing the number of digital artists.
2. Medical and Biological Innovation: AI advancements in protein design are paving the way for breakthroughs in medicine, such as creating proteins to target specific diseases, thanks to deep learning models that understand and generate biological structures.
3. Engineering Efficiency and GPU Constraints: AI development requires extensive computational resources. Chinese firms like Kai-Fu Lee’s 01.AI are innovating under constraints by engineering highly efficient systems, using a fraction of the GPU power needed by larger competitors like OpenAI.
4. Enhanced Work Productivity and Personal AI Agents: AI will fundamentally change productivity, creating specialized personal AI agents for tasks like cybersecurity, customer service, and marketing. Work may increasingly focus on managing and training AI systems rather than manual tasks.
5. Future Skill Recommendations: For young people entering the workforce, leaders have varied advice: learn AI tools, consider programming for foundational knowledge, and follow personal passions. Programming and AI knowledge will be vital, though the primary “code” of the future may shift to human languages like English.

How To Build The Future: Sam Altman — Y-Combinator interview

1. Conviction in AGI and Long-Term Vision: Sam Altman’s journey with OpenAI shows his early and focused commitment to artificial general intelligence (AGI), a conviction that drove the founding of OpenAI despite industry skepticism.
2. Scaling and Abundance Potential: Altman emphasizes the potential of abundant intelligence and energy to unlock significant technological advancements, including applications in climate, space exploration, and possibly new physics discoveries.
3. Early Career Lessons and Platform Shifts: From founding Loopt to OpenAI, Altman’s career reflects learning from platform shifts (like mobile to AI), underscoring how timing, market understanding, and adaptability are crucial for tech success.
4. Role of High-Conviction Startup Culture: Altman advocates for startups betting on high-conviction ideas, as large companies may miss rapid shifts in AI’s potential; he encourages founders to focus on bold bets rather than safe, incremental improvements.
5. Advice for AI Founders: Altman advises building something valuable with AI’s accelerating capabilities while remembering that traditional business principles — such as creating a moat and providing real value — still apply.

Y Combinator CEO Shares How They Pick Winners, Advice For Founders + Lessons From Paul Graham

1. Look for High Potential with Simple Processes: Y Combinator evaluates thousands of applications quickly, focusing on passion, technical talent, and the ability to adapt. They often invest early — even pre-idea — and rely on a straightforward 10-minute interview.
2. Build Something People Want: Y Combinator’s mantra, inspired by Paul Graham, emphasizes creating genuine solutions for real needs. Founders should prioritize understanding customers’ problems directly and iteratively improve their products.
3. Prioritize Customer-Centric Development: YC emphasizes building a worldview directly informed by customers’ needs. Founders are advised to engage deeply with real users, observe their workflows, and continuously iterate based on direct feedback to ensure the product remains valuable and relevant.
4. Leverage Small, Agile Teams: YC encourages founders to start with “two-pizza teams” (small enough to be fed with two pizzas), especially for AI-driven businesses. This keeps startups nimble and prepared to scale without the overhead of large teams.
5. Play Your Own Game: Founders are advised to avoid framing their work based on competitors or investor expectations. Focus on serving customers and building a strong, differentiated value proposition that stays true to the startup’s vision.

Check out the week’s top AI Videos on the AI-Essential Vids section of the INSEAD AI YouTube Channel which is regularly updated.

For a broad selection of AI videos, check out the INSEAD AI YouTube Channel (regularly updated)

Past editions — AI NEWS YOU MISSED

View list

SUBSCRIBE TO GET UPDATES IN YOUR MAILBOX

Follow Robert Maciejko on LinkedIn or X (Twitter)