Funston Brief (Issue No. 26)
This month, our featured commentary discusses foundation models applied to robotics. We also dive into the business models of AI companies and discuss where the VC industry is headed.
Foundation models and industrial robots
Covariant has unveiled RFM-1, an 8 billion parameter transformer model designed to revolutionize industrial robotics. Trained on a diverse dataset including text, images, videos, robot actions, and sensor readings, RFM-1 aims to simplify robot operation akin to prompting language models for text generation. The dataset, derived from Covariant's warehouse deployments worldwide, exposes RFM-1 to various real-world scenarios, enhancing its understanding of the physical environment.
RFM-1 enables intuitive human-like interactions with robots, allowing users to issue instructions in plain English and facilitating robots to articulate issues to humans. However, RFM-1 remains a prototype with limitations in resolution and frame rate, hindering its ability to effectively model small objects and rapid motions.
Despite being in the prototype stage, RFM-1 signifies a significant advancement in robotics, potentially accelerating the field's progress. RFM-1 promises to unlock new capabilities and usher in a new era of agile and intelligent industrial automation systems by integrating advanced AI techniques into existing industrial robots, such as data collectors and control systems.
ACTIONABLE INSIGHTS.
The most important AI model is the business model: With the explosion of GenAI and foundation models, there is a growing debate about the most important AI model in the months and years ahead. Countless models with varying degrees of “openness” are vying to supplant GPT-4, including models from Meta, Amazon, Google, Microsoft, Anthropic, and Mistral. But, these debates are missing the key point. The most important model required to remain an enduring winner in AI is a compelling business model.
Key Insight: As AI models proliferate, companies across the AI stack will need to think deeply about their business models in general and their pricing and packaging strategies to ensure long-term success. There is tension in AI business models between achieving near-term scale and delivering strong unit economics over time. This tension will become more acute as incumbents bundle AI with existing software, and customers demand a clear ROI for the incremental dollars they are paying for AI to improve business processes or team productivity. Link
CEO excellence: How do leaders assess their performance? McKinsey published some early results from their proprietary CEO Excellence Assessment Tool, offering insights into the key issues facing more than 100 CEOs and where they feel most confident and vulnerable.
Key Insight: Even at this preliminary stage, the data set is already yielding rich and often counterintuitive insights into the challenges and uncertainties faced by CEOs. McKinsey finds, for example, that CEOs generally feel confident about their ability to manage their own personal effectiveness but can struggle to manage the board and engage with key stakeholders. Link
EAR TO THE GROUND.
Podcast Episode: AI, Crypto, 1000 Elon Musks, Regrets, Vulnerabilities, & Managerial Revolution
Guests: Marc Andreessen
Takeaway:
Dwarkesh Patel chats with Marc Andreessen in an exciting exchange covering some cool stuff: how venture capital brings back bits of old-school capitalism in today's mostly corporate world, what sets apart great startup founders from seasoned execs, the nitty-gritty of valuing startups early on, and more.
“When you're doing new things [...], there's this role in the background who's like - Okay, this one, not that one. This founder, not that founder. And those people play a judgment and taste role; they play an endorsement, branding, and marketing role. And then they often play a financing role. And they often are very hands-on, and they try to contribute to the success of the project. So what I'm pretty confident about is there will be somebody like us who is doing that [very fundamental role] in 50 years, 100 years, 200 years.”
- Marc Andreessen, co-founder of A16Z, on the future of venture capital
$18.4 billion
NVIDIA's AI chip sales skyrocketed to $18.4 billion in Q4 2023, up from $3.3 billion a year prior. Cloud giants fueled the surge, with one major player (likely Microsoft) accounting for 19% of sales. Analysts debate NVIDIA's long-term lead, but for now, they dominate a supply-constrained market in a cyclical industry. Link
PROFILE.
In a feature for The New Yorker, Stephen Witt explores the world of NVIDIA's CEO, Jensen Huang. Dubbed the man of the moment, perhaps even of the entire decade, Huang's leadership has propelled NVIDIA's supercomputing GPUs into unparalleled prominence within the tech industry. Yet, with mainstream media coverage of tech figures, one can't help but wonder: Is this the peak, or is there more to come? Enjoy the read!
“Deep learning is not an algorithm. Deep learning is a method. It’s a new way of developing software.” - Jensen Huang, co-founder of NVIDIA.
ON MY RADAR.
Apple fortified iMessage with quantum-proof encryption, protecting it from super-powered computers that crack current codes. Link