The AI Rundown by Lightscape Partners – 03/25/24

Good morning, and welcome back to another edition of The AI Rundown.

Last week in AI,

• NVIDIA hosted their flagship AI conference, announcing key breakthroughs in chip design and robotics.

• Stability AI continues to be anything but stable, as the founding CEO announces his departure.

• Insiders at OpenAI leaked details about the GPT-5 timeline and new capabilities.

• Microsoft’s AI team gets even stronger as the Inflection co-founder announced he is joining the team.

Check out the AI conferences happening this month here.

If you’d like a quick refresher on industry terms and the current landscape of AI, skip to the bottom.

Top AI Stories of the Week

Stability AI’s CEO steps down amid controversy to pursue decentralized AI. Link.

• The Stability AI founder and CEO Emad Mostaque resigned following 3 of the company’s researchers who developed the original Stable Diffusion algorithm left.

• One of the major investors, Coatue Management, had been calling for Mostaque to step down.

• Mostaque stated he stepped down to pursue decentralized AI.

GPT-5 features and capabilities and timeline revealed in latest OpenAI leak. Link.

• In new leaks from internals at OpenAI, GPT-5 is allegedly set to release sometime this Summer.

• The new GPT may have the ability to “call AI agents” developed by the community.

Microsoft hires Inflection AI’s co-founder to lead their AI push. Link.

• Microsoft hired Mustafa Suleyman to oversee the tech company’s consumer AI product development.

• Suleyman will serve as executive VP of Microsoft AI and report directly to Satya Nadella.

• The Inflection AI co-founder announced his next journey in a post on X.

Software + Hardware

Intel to receive $20 billion in AI chip incentives from the Biden administration. Link.

• Intel will receive almost $20 billion in the form of grants and loans to subsidize innovation chip production.

• Some of the funding will go into two new factories to be built in Arizona and the renovation of a third.

Researchers at Stanford give AI the ability to have an inner monologue. Link.

• Researchers created an AI model that will “think” internally before generating answers, leading to better and more accurate responses.

• The team built their model on Mistral 7B, an open-source LLM.

Apple may let Google power iPhone AI features. Link.

• Apple and Google are in discussion to allow Gemini to power iPhone AI features.

• Apple would license Gemini to implement new AI specific features on the iPhone.

• The strategic partnership allows Apple to build an AI edge amid rumors the tech company is falling behind.

Venture

Nvidia in talks to acquire Run:ai for $1bn+. Link.

• The Israeli startup was founded in 2018 and has developed a “orchestration and virtualization” layer for AI workloads running on GPUs.

• Run:ai last raised $75 million in March 2022.

NVIDIA GTC 2024

Blackwell, the AI “superchip” that reduces cost and energy consumption by upwards of 25X. Link.

• Nvidia announced Blackwell B200 GPU, the “world’s most powerful chip” for AI.

• The chip is purpose built to handle the first trillion-parameter AI model.

• The B200 GPU offers up to 20 petaflops of FP4 power.

• Nvidia claims 2,000 Blackwell GPUs can train a 1.8 trillion parameter model with only four megawatts of power.

Project GR00T “General Robot 00 Technology”: A foundational model purpose-built for humanoid robot learning. Link.

• The foundational model will allow for humanoid robots to understand spoken language, mirror human movements, and better develop “skills”.

• Robotics companies partnering with Nvidia include Figure, Apptronik, Agility Robotics, Sanctuary AI, and Unitree.

Upcoming AI Conferences

IMAGINE AI LIVE ‘24

• Fontainebleau, Las Vegas, NV

• March 27 – 28, 2024

Get up to speed on the current landscape

• AI Term Glossary

• Current AI Landscape from Sequoia Capital

Thank you for reading the AI Rundown by Lightscape Partners. Please send any questions, comments, or suggestions to [email protected].