Long Take: Trusting Artificial Intelligence based on ZK Proofs, and the $10B fraud market
Long Take: Trusting Artificial Intelligence based on ZK Proofs, and the $10B fraud market
And a look at the recent controversy regarding Bridgewater Associates
Gm Fintech Architects —
Today we are diving into the following topics:
Summary: Trust in the financial sector, as exemplified by Bridgewater Associates' claims about its investment processes, is a critical yet often elusive asset. Despite Bridgewater being the largest hedge fund, skepticism surrounds its performance and strategies, highlighting the complex nature of trust in finance. In payments, trust is frequently compromised, evident in the annual $10-15 billion losses due to card fraud. Zero Knowledge Proofs (ZKPs) are emerging as pivotal tools in reinforcing trust, with ZKPs ensuring transaction authenticity without revealing sensitive details. ZKML solutions, in particular, are instrumental in verifying the outputs of AI models, enhancing the reliability and integrity of financial systems.
Topics: artificial intelligence, zero knowledge proofs, blockchain protocols, investment management, fraud, identity
If you got value from this article, let us know your thoughts! And to recommend new topics and companies for coverage, leave a note below.
Long Take
Why Trust Matters
Spend enough time in certain corners of the Internet, and you’ll hear about the importance of programmable trust, decentralization, and resistance to adversarial attacks on your code. There’s a certain dystopian Neuromancer flavor to this, as if the nation states have already launched their AI gods into cyberspace to manipulate us mortals into obedience.
It’s weird stuff to the uninitiated.
But we don’t have to start there. Instead, we can build an intuition from things we already know.
Take for example, Bridgewater Associates. We are fans of Ray Dalio, in the same way we admire Elon Musk or Jeff Bezos or Sam Altman or Eliezer Yudkowsky — squinting our eyes against the sun and covering the face from the absurdity of it all, to notice the exceptional progress underneath. There is a chance to both go blind, as most did with Sam Bankman-Fried, and a chance to notice something profound.
A new book by New York Times finance reporter Rob Copeland called “The Fund” paints Bridgewater in negative terms. We refer to this extended article by the author to highlight a few key data points. Despite underperforming the markets, Bridgewater is still the largest hedge fund according to Pensions & Investments.
Copeland tries to ask the question about how the firm makes money. On the surface, Bridgewater markets itself as a highly sophisticated quantitative investment firm, integrating massive datasets across macroeconomics and fundamentals to generate alpha. Yet given its returns profile and lack of positions in markets, various investigators and whistle-blowers thought of it as a Bernie Madoff-type scheme. After some digging, Copeland makes the following point:
In the beginning, Dalio’s approach was to use deterministic rules that govern macro-economics to make bets in the markets. These rules could be stored in spreadsheets or the minds of employees. Today, however, enormous machine learning clusters can process anything and everything there is to know about financial data, and simple rules lead to underperformance.
There is a difference between making the claim that you run the most sophisticated quant fund in the world, with thousands of statistically significant investment rules, and actually running one. As an asset allocator, you would probably want to know whether the claims being made to you about a decision-making process are true or false.
Whether you believe the claims, requires something called trust.
Trust is very squishy. People abuse it constantly. Further, signaling trustworthy features — like wearing suits, having an advanced degree, speaking using formal language, being tall — is often confused for being trustworthy. Politicians, the people who are most acutely in the trust business, end up being some of the least trustworthy as a result. Depressingly, focusing on signal is more advantageous when relating to people than focusing on results.
Now imagine if you had a magic algorithm to verify whether what Ray Dalio and Bridgewater were representing about the operation of the investment process was true or false.
Finance, Fintech, and Identity
The Bridgewater example is anecdotal and perhaps insufficiently persuasive.
Let’s flip this instead into quantifying the lack of trust in financial services. What is endemic, highly scalable, and experienced by all financial participants?
Fraud.
In the ideal scenario, everyone participating in a financial transaction is who they say they are, and they are using their own financial resources and products. Yet in payments alone, card fraud losses per year range between $10B and $15B per year, with 5% of consumers affected, or over 10 million people.
The abuse of the presumption of trust — for example, a merchant being willing to accept a stolen card — is theft.