The impact of the AI revolution on asset management

Michael Kopp, Michael Kopp 0001arXiv 2023 · DOI · arXiv

Introduction

The landscape of asset management is undergoing a fundamental shift. While algorithmic trading has been a staple of financial markets for decades, it has primarily focused on high-frequency execution or rigid, human-designed rules. Complex "fundamental" trading—tasks requiring the synthesis of unstructured data like legal filings, news, and geopolitical events—has remained the exclusive domain of human intuition. However, recent breakthroughs in deep learning (DL) are challenging this status quo. This paper explores how AI is evolving from a mere tool into an autonomous agent capable of reasoning and strategy discovery, potentially disrupting traditional asset management models.

The Nature of the Deep Learning Revolution

The current AI revolution is distinct from previous technological shifts. It is characterized by the emergence of Large Language Models (LLMs) and foundation models that can process "dirty," raw, internet-scale data without the need for manual labeling. This allows systems to "understand" market dynamics in ways previously impossible for quantitative models.

The paper's structure highlights the transition from historical context to a new framework for AI-driven asset management.

As noted in the paper's structure, the "odd nature" of this revolution involves models that can perform "zero-shot" reasoning. For instance, an LLM might deduce the ripple effects of a Suez Canal blockage on specific shipping routes and raw materials faster than a human analyst. Yet, these models also exhibit "alien" behaviors—hallucinations or counter-intuitive patterns that challenge human expectations.

An AI-generated image exhibiting structural artifacts, illustrating the 'odd nature' and 'black box' characteristics of deep learning models.

A Framework for AI Maturity

To assess how funds are integrating these technologies, author Michael Kopp proposes a four-level maturity framework:

  • Level 0: Human-Led. Decisions are made by humans, supported by traditional tools like Bloomberg terminals.
  • Level 1: Automated Execution. The strategy is human-designed and validated, but execution is automated.
  • Level 2: Automated Strategy Discovery. AI identifies potential strategies, but a "human in the loop" provides supervision and final validation.
  • Level 3: Full Autonomy. The system autonomously handles strategy discovery, capital allocation, and risk management.

The paper argues that Level 3 funds are no longer a theoretical future but a technological possibility. Such funds could operate with "super-human intuition," discovering trading patterns that are mathematically sound but may seem "alien" to human traders, much like the moves made by AlphaGo.

Key Insights: The Black Box and Regulatory Hurdles

A significant barrier to AI adoption is the "black box" problem. The author posits that we should treat AI traders similarly to human ones. Human biological neural networks are also black boxes; we judge them by their performance and verbal justifications, not by auditing their neurons. Similarly, AI should be evaluated based on risk parameters and performance rather than line-by-line code explainability.

Furthermore, the regulatory environment remains a hurdle. Traditional frameworks are ill-equipped for autonomous, self-evolving strategies. Consequently, Level 3 funds are likely to emerge first in less regulated spaces, such as cryptocurrency markets, before moving into mainstream finance.

Conclusion

The convergence of big data and massive compute power has brought Level 3 asset management within reach. Funds remaining at Level 0 or 1 face significant disruption risk as autonomous systems scale their ability to exploit market inefficiencies. While the field still lacks a formal "theory" of deep learning architecture, the practical feasibility of autonomous financial agents marks a turning point for the industry.

Summary updated

Abstract

Recent progress in deep learning, a special form of machine learning, has led to remarkable capabilities machines can now be endowed with: they can read and understand free flowing text, reason and bargain with human counterparts, translate texts between languages, learn how to take decisions to maximize certain outcomes, etc. Today, machines have revolutionized the detection of cancer, the prediction of protein structures, the design of drugs, the control of nuclear fusion reactors etc. Although these capabilities are still in their infancy, it seems clear that their continued refinement and application will result in a technological impact on nearly all social and economic areas of human activity, the likes of which we have not seen before. In this article, I will share my view as to how AI will likely impact asset management in general and I will provide a mental framework that will equip readers with a simple criterion to assess whether and to what degree a given fund really exploits deep learning and whether a large disruption risk from deep learning exist.
BibTeX
@misc{kopp2023impact,
  author = {Michael Kopp and Michael Kopp},
  title = {{The impact of the AI revolution on asset management}},
  year = {2023},
  eprint = {2304.10212},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2304.10212},
}
Figure