This article is part of Fintech Leaders, a newsletter with over 60,000 builders, entrepreneurs, investors, regulators, and students of financial services. I invite you to share and sign up! Also, if you enjoy this conversation, please consider leaving a review on Apple Podcasts, Spotify, or wherever you get your shows so more people can learn from it.
I sit down with David Magerman, Managing Partner and Co-Founder at Differential Ventures, an early-stage fund that invests in data-focused entrepreneurs building enterprise technology.
In a past life, he spent 22 years in many leadership roles at Renaissance Technologies (RenTech), a hedge fund with over $100 billion in AUM and probably the most successful fund in history.
David is also a contrarian thinker and a non-conformist, which made our conversation even more interesting.
In this episode, we discuss:
How David and a small group of engineers helped introduce quantitative data science to trading and public capital markets investing
David has been at the forefront of computer science since the 80s and throughout his career has been deeply involved in early efforts to apply machine learning and data analysis technologies to capital markets investing. Recognizing the potential of data science in a field ripe with data, Magerman and some of his IBM colleagues joined Renaissance Technologies in the mid-90s. This career move marked a significant pivot. At Renaissance, a quantitative hedge fund known for its secretive and successful trading strategies, David and his team applied their data science expertise to successfully develop algorithms that could predict market movements. This transition underscored a novel application of data science, moving from theoretical research to the practical, high-impact world of trading, where accuracy and rigor in research directly influenced financial outcomes.
A key to their success was the introduction of rigorous computer science principles to the development of trading systems. David highlighted the importance of engineering discipline, brought by the IBM team, in building sustainable, testable, and flexible software systems. This approach allowed for more effective application of research findings to trading strategies, significantly enhancing the performance and effectiveness of the fund's results and operations.
Why he is highly skeptical of the new wave of GenAI, particularly in fintech
“I think that we're going to discover that we're applying MLMs and generative AI, to too many problems, I, my guess, is somewhere around like 5% of the problems we're applying to, it's actually capable of solving well, and the other 95% are going to be a sinkhole of wasted resources, opportunity, time, money.”
Overestimation of AI's Capabilities: David argues there is currently a fundamental misinterpretation of what large language models (LLMs) are currently capable of. Despite the appearance of advanced intelligence, these models primarily excel at memorizing and reiterating data rather than demonstrating genuine understanding or innovation. This overestimation of AI's capabilities could lead startups to invest heavily in solutions that are not as revolutionary or effective as anticipated, potentially resulting in significant resource wastage.
Misalignment with Problem-Solving: According to Magerman, only a small fraction of problems, possibly around 5%, are genuinely solvable by current GenAI technologies. The remaining 95% could turn into a "sinkhole" of wasted resources, time, and money. This stark discrepancy underscores the importance of critically assessing whether GenAI is the right tool for the challenges a startup aims to address, rather than applying it as a one-size-fits-all solution.
Risk of Public Failure and Customer Backlash: Founders should be warned of the potential for high-profile failures within the AI space that could lead to a significant loss of customer trust and interest, not just for the companies directly involved but for the entire sector. Startups relying heavily on GenAI must prepare for the fallout from such events, ensuring they have strategies in place to retain customer confidence and demonstrate the distinct value and reliability of their offerings.
Importance of Proven Deployment: For VC investors, the proof of a product's viability should come not from pilot projects or proofs of concept but from sustained deployment in the market. David expressed a preference for investing in companies whose products have been actively used and validated by customers over significant periods. This approach suggests that startups should focus on achieving and demonstrating real-world effectiveness and customer satisfaction, rather than relying solely on the potential of GenAI technologies.
The reasons why you cannot apply data science analysis to venture capital investing
"I don't think it makes sense to apply data science to, to venture investing."
Lack of Immediate Feedback: Unlike quantitative trading, where immediate financial outcomes signal the effectiveness of a strategy on a daily basis, venture capital operates on a much longer timescale, often spanning three to ten years. This extended horizon makes it challenging to apply data science effectively, as the immediate feedback necessary to iterate and refine models in real-time is absent. In venture capital, the success or failure of an investment may not be apparent for years, complicating the application of data-driven decision-making.
The Role of Intuition and Human Oversight: Magerman reminds us of the importance of intuition and human oversight in venture capital investing. While data science can enhance efficiency and support the investment process, it cannot replace the nuanced understanding and judgment that experienced investors bring to the table. The complexity of evaluating startups, including assessing team quality, market potential, and technological innovation, requires a depth of analysis beyond what current data science methods can offer.
The Challenge of Applying the Law of Large Numbers: The statistical principle that allows quantitative trading strategies to be validated through large volumes of transactions over short periods does not apply well to venture capital. Given the long investment horizons and the relatively small number of investments a firm might make, there isn't enough data to conclusively prove the effectiveness of a data-driven investment strategy in the venture capital context.
A compelling argument to introduce friction in online communication processes to restore balance and reduce its negative effects on society… and a lot more!
“I'm deeply concerned about the state of information technology in general, in the world, when it comes to news, media, social media, and just and the surveillance economy that we have now.”
Surveillance Economy and Privacy Invasion: Magerman points to the alarming capabilities of modern technology to track and store personal location data from vehicles, illustrating the invasive nature of the surveillance economy. This capability, coupled with the illusion of anonymization, poses significant risks to individual privacy, as seemingly anonymous data can often be re-associated with specific individuals based on their patterns of movement.
Information Overload vs. Knowledge: Despite the exponential growth in sources of information, David argues that the quality and veracity of knowledge have declined. The proliferation of news outlets, social media platforms, and other digital content has not led to a better-informed public. Instead, it has created an environment where misinformation spreads rapidly, and the effort required to verify the accuracy of information often outweighs the benefits of consuming it in the first place.
Erosion of Public Discourse and Decision Making: The current state of online communications has complicated the ability of governments to function effectively and has influenced elections and public policy in ways that may not align with the long-term interests of society. The sheer volume of information and the mechanisms of its dissemination have contributed to a reduction in the public's attention span, making it increasingly difficult to engage in meaningful discourse or concentrate on complex issues.
Proposed Solutions to Restore Balance: To counteract these challenges, David Magerman suggests introducing friction into digital communications to slow down the pace at which information is disseminated and consumed, bringing it closer to human speed. One innovative idea he proposes is a payment system for receiving emails and messages, which would impose a cost on senders and potentially reduce the volume of unwanted communications. This approach aims to recalibrate the balance between the speed of computer-generated communications and the human capacity to process information, thereby mitigating some of the negative impacts of the digital age on attention spans and privacy.
David’s favorite book authors: Jonathan Henry Sacks, Isaac Asimov, and J. R. R. Tolkien
Episode Chapters
00:00 Applying LLMs and Generative AI to a Limited Number of Problems
03:02 David Megerman's Background and Transition to Venture Capital
07:52 The Challenges of Research and Product Development at IBM
10:47 Applying Data Science to Capital Markets Analysis at Renaissance Technologies
14:11 The Transition from Public Capital Markets to Venture Capital
16:10 The Limitations of Data Science in Venture Investing
17:38 Concerns about the New Wave of Generative AI
21:31 The Importance of Recognizing the Limitations of LLMs and Generative AI
23:57 The Dangers of Misleading Demos and Overreliance on LLMs
26:23 Use Cases for LLMs in Fintech and the Importance of Human Oversight
28:47 Concerns about Information Technology and the State of News and Media
33:39 Introducing Friction in Information Communication Processes
38:02 Investing in B2B Solutions and Avoiding Problems in Consumer Products
40:28 Recommended Books: Lord Rabbi Jonathan Sacks and Classic Science Fiction
Want more podcast episodes? Join me and follow Fintech Leaders today on Apple, Spotify, or your favorite podcast app for weekly conversations with today’s global leaders that will dominate the 21st century in fintech, business, and beyond.
Share this post