This episode dives into the hidden economics of GPU scarcity—how AI and crypto are colliding over compute power, and what this means for investors.
Yanez MIID: Adversarial Data for Financial Crime Prevention
- Jose Caldera introduces Yanez MIID (Multimodal Inorganic Identity Data Sets), Bittensor Subnet 54, which operates a network generating synthetic identities to prevent financial crime. The subnet's core function is to create "adversarial data" that stress-tests Know Your Customer (KYC) and Anti-Money Laundering (AML) systems used by financial institutions. This helps banks ensure their systems effectively catch fraud, money laundering, and sanctions evasion, meeting stringent regulatory requirements and avoiding significant fines. Jose highlights the immense scale of the problem: "350 billion in the last 15 years" in fines, $450 billion lost annually to fraud in the U.S., and an estimated $2 trillion laundered globally each year.
- Strategic Implication: The demand for robust financial crime prevention tools is massive and growing, driven by regulatory pressure and escalating fraud. Solutions leveraging decentralized adversarial networks like Yanez MIID offer a novel approach to system testing and improvement.
Crypto vs. Traditional Finance in Money Laundering
- Mark Jeffrey raises the common perception of crypto as a haven for illicit activities, often cited by figures like Elizabeth Warren. Jose Caldera counters this, explaining that blockchain technology, by its very nature, is often easier to track than traditional cash transactions. Tools like Chainalysis and Elliptic are highly effective in tracing nefarious actors on a blockchain. He asserts that crypto represents only a "tiny little fraction" of the global money laundering problem, which is predominantly facilitated by the traditional financial system.
- Actionable Insight: Crypto's inherent transparency, often misconstrued as a weakness, is a powerful tool for law enforcement and financial intelligence. Investors should recognize that blockchain analysis capabilities make crypto less appealing for large-scale, untraceable money laundering compared to traditional methods.
The Adversarial Attack Contest on Bittensor
- Subnet 54's miners are engaged in a unique contest: to create new, sophisticated vectors of attack against identity verification systems. Initially focused on simple name variations for sanction screening, the network has evolved to challenge systems with novel attack methods. Jose explains that while 70-75% of miner submissions are attempts to "game the system," a significant 25% represent "actual attempts to break these things," uncovering new vulnerabilities. This decentralized approach generates an astonishing volume of attacks, potentially "20,000 different ways every day," far surpassing the capabilities of traditional in-house QA teams or ethical hackers.
- Strategic Implication: Decentralized adversarial networks on Bittensor offer an unparalleled, scalable, and creative testing ground for AI models and security systems. This model can uncover "zero-day attacks" (previously unknown vulnerabilities) at a speed and scale impossible with conventional methods, providing a significant competitive advantage for security solutions.
Automating Attack Evaluation and Market Opportunity
- Evaluating the high volume of creative attacks from miners presents a challenge. Yanez MIID employs a multi-layered post-processing system: initial layers are automated to discern attack characteristics, while final layers involve manual evaluation or sampling due to the complexity and cost of calling advanced models. The market for financial crime prevention is substantial, with approximately $200 billion spent annually, split between human resources ($120 billion) and technology ($80 billion). Yanez MIID targets large, multi-billion dollar financial institutions and global organizations, not small fintechs, due to the significant value proposition. Their primary use case involves automating the monitoring of sanction lists, a task that can consume hours daily for banks operating in multiple jurisdictions. Jose notes remarkable early success, securing multi-billion dollar clients within their first year, a feat he describes as "gigantic" compared to his previous ventures.
- Actionable Insight: The financial crime prevention market is ripe for AI-driven automation, particularly in areas like compliance and identity verification. Solutions that can demonstrate clear ROI by reducing manual labor and enhancing detection capabilities will capture significant market share, especially those with a unique, scalable data generation model.
TFlow Impact and Strategic Adjustments
- Despite early success, Yanez MIID faced a challenge with Bittensor's TFlow mechanism, which caused their emissions to drop to zero. Jose, with nearly 30 years of entrepreneurial experience, acknowledges that TFlow's concept "makes a lot of sense" but highlights a misalignment with their current revenue model. Their multi-year enterprise contracts typically provide revenue annually, which doesn't align with TFlow's 86.8-day moving average for measuring "excitement" and liquidity pool activity. As a startup, Yanez MIID, like many others (e.g., Uber, Amazon), is currently burning capital. To address this, they are "repackaging a product line" to generate a more consistent, stream-like revenue, aiming for a Q1 launch.
- Strategic Implication: Bittensor subnet owners must critically evaluate and adapt their business models to align with the protocol's dynamic incentive mechanisms like TFlow. Investors should scrutinize a subnet's revenue generation strategy and its compatibility with Bittensor's economic cycles, favoring those with agile product packaging and diversified revenue streams.
Bittensor's Experimental Nature and Future Outlook
- Mark and Jose discuss Bittensor as an experimental "petri dish" where continuous adjustments are expected from the protocol, subnet owners, and investors alike. Jose emphasizes the need for adaptability, stating, "I went into this understanding that there were things that we needed to tweak and things that we need to adjust." He praises the thoughtfulness behind Bittensor's changes, even if their implementation requires further refinement. Jose highlights the "unnatural advantage" Bittensor provides: a "giant army" of creative, adversarial attackers working around the clock, a model akin to Bitcoin's decentralized mining. This offers an unprecedented scale of creativity and attack vectors that traditional companies cannot replicate, despite the need to manage data privacy and enterprise client expectations.
- Actionable Insight: Bittensor represents a frontier in decentralized AI, demanding a high tolerance for experimentation and rapid adaptation from all participants. Subnet owners who can leverage the protocol's unique incentive structures to build scalable, adversarial networks gain a significant, almost insurmountable, competitive edge in their respective markets.
The Rise of Deepfakes and Biometric Security
- Jose distinguishes between document capture for convenience (e.g., dealing with glare or cropping) and for security. He identifies deepfakes as the "true thing that is going to change how this world verification identity operates fundamentally." Deepfakes break the "something you are" principle of security, which relies on biometrics. This threat extends beyond faces to voice and video, fundamentally challenging every system reliant on biometric verification, from financial institutions to border security. Yanez MIID's future direction is to focus on producing data to test and train systems to detect these increasingly sophisticated deepfakes. Jose notes that while the market need is immense and customer receptivity is high, the sales cycle for multi-billion dollar clients can be lengthy, often taking nine months or more to navigate procurement and security processes.
- Strategic Implication: Deepfake detection is rapidly becoming a critical, high-value segment within identity verification and cybersecurity. Investors and researchers should prioritize solutions that can effectively generate and utilize adversarial data to train and test deepfake detection models, as this technology will underpin future security infrastructure.
Reflective and Strategic Conclusion
- Yanez MIID exemplifies how Bittensor's decentralized adversarial networks offer an unparalleled advantage in combating financial crime, particularly against the rising threat of deepfakes. Investors and researchers should track how subnets adapt their revenue models to Bittensor's TFlow mechanism and focus on projects leveraging decentralized AI for critical, high-stakes security applications.
