AI Scaling: Reaching a Potential Plateau?

Ashu Garg, in "Has AI scaling hit a limit?", examines the dominant belief that increasing the scale of AI models (bigger networks, more data, more computing power) leads to greater intelligence. While acknowledging the immense success of this approach, exemplified by ChatGPT's rapid advancements, he questions if we are approaching a limit.

Garg argues that current Large Language Models (LLMs) have achieved "minimum viable intelligence", driving improvements across industries and enabling a new wave of AI-native products. However, focusing solely on scale might limit us to "local maximum" in AI development. He suggests that the field requires a shift towards fundamentally different models, echoing Ilya Sutskever’s call for renewed “wonder and discovery.”

Venture Capital: The Long Road to Unicorn Status

Data from Ilya Strebulaev on LinkedIn reveals the funding journey of nearly 2,000 US unicorns. Key findings include:

The average unicorn takes 5 funding rounds.
While some achieve unicorn status in less than 2 rounds, the "sweet spot" is 3-5 rounds.
50% of unicorns require more than 5 rounds, with some needing as many as 18.

This data highlights the long and arduous path to unicorn status, emphasizing that consistent funding is crucial for startup success. However, commenters raise valid concerns about the 27-year timeframe of the data, suggesting a need for analysis of more recent trends.

Pitching to VCs: Rethinking Market Sizing

Ivan Landabaso's LinkedIn post advocates replacing the traditional market sizing slide with an “idea maze” slide, as proposed by Balaji Srinivasan. This slide should:

Offer insightful analysis of the market's past, present, and future.
Showcase the founder's deep understanding of potential pitfalls and challenges.
Proactively address anticipated questions and solutions to hurdles.
Avoid jargon and personalize the pitch for the specific audience.

By replacing a static market size number with a dynamic picture of market understanding, founders can demonstrate their strategic thinking and adaptability, crucial qualities for investors.

Marketplace Businesses: Captivity and Revenue Quality

Rick Zullo, in "Marketplaces, Brokers and Storefronts", emphasizes the power of market "captivity" for marketplaces like eBay. Despite a poor user experience (UX), eBay thrives due to its dominance in the collectibles market:

"Make no mistake, eBay is a terrible product (at least from a UX perspective), but its captivity over the collectibles market results in it selling over $10b in collectibles each year despite exorbitant fees)."

This observation highlights that “all revenue is not created equal,” as famously stated by Bill Gurley. Companies with strong market captivity can generate consistent revenue despite shortcomings, emphasizing the importance of identifying businesses with durable competitive advantages.

European Venture Capital: Challenges and Opportunities

The "EIF VC Survey 2024" paints a mixed picture of the European VC landscape. While fundraising sentiment has reached an all-time high, challenges remain in areas like:

Competition for deals: Competition has increased, creating a more challenging environment for VCs to secure promising investments.
Difficult exit environment: Exits remain challenging, with exit prices decreasing for the third consecutive year. IPOs, in particular, have fallen to an all-time low.
Scale-up financing gap: A lack of sufficient scale-up financing hinders the growth of promising companies and limits successful exits.

However, the survey also highlights promising opportunities:

Strong VC confidence: Despite challenges, VCs remain confident about the future of the European VC market.
Attractive investment sectors: AI and energy/climate are identified as particularly promising sectors for future investments.
Promising investment countries: Germany, the UK, and France are ranked as the most promising European countries for VC investments.

Beyond the Scaling Laws

Agora - The Marketplace of Ideas