My third question was – “Accepting the ‘black swan’ exception how much of these other ‘human essential’ activities could an advanced Artificial Intelligence take over to bring the markets to as close to 100% AI run as possible?”
This was where Gemini started to get into speculation:
Gemini came up with this:
That’s an excellent follow-up question! Accepting the “black swan” limitation (the inability of AI, currently, to handle truly unprecedented and fundamentally new events with human-like judgment), let’s explore how much more AI could take over to bring financial markets as close to 100% AI-run as possible.
It’s important to differentiate between automation of existing processes and truly autonomous decision-making in novel circumstances. AI is rapidly excelling at the former, but the latter remains a significant challenge.
Here’s a breakdown of areas where advanced AI could theoretically take over much more, along with the inherent challenges and limitations:
Areas Where AI Could Take Over Significantly (Closer to 100%):
-
Algorithmic Trading & Execution (Already very high, could become near-total):
- Sentiment and News Trading: Advanced NLP and generative AI could continuously monitor all global news, social media, and regulatory announcements, extracting nuanced sentiment and fundamental signals to drive trading decisions instantly, without human review.
- Automated Strategy Development: AI could be tasked with generating and testing novel trading strategies, learning from their performance and adapting them in real-time, requiring minimal human intervention once the meta-learning system is established.
-
Risk Management (Vastly enhanced, potentially near-autonomous for known risks):
- Real-time Risk Monitoring & Mitigation: AI could continuously monitor market, credit, operational, and liquidity risks across all entities, automatically adjusting positions, initiating hedges, or even freezing certain activities if predefined risk thresholds are breached.
- Stress Testing & Scenario Analysis: AI could generate and run thousands of complex stress test scenarios in real-time, identifying vulnerabilities and recommending preventative measures.
- Fraud Detection & Anti-Money Laundering (AML): AI is already very strong here. It could move towards proactive identification of illicit activities in the financial markets and even automated reporting to regulatory bodies.
-
Compliance & Regulatory Reporting (Significant automation potential):
- Real-time Compliance Monitoring: AI could continuously monitor all transactions and activities for adherence to existing regulations, flagging or automatically correcting deviations.
- Automated Reporting: Generative AI could draft and submit complex regulatory reports, ensuring accuracy and timeliness based on collected data.
- Regulation Interpretation: AI could analyse new and changing regulations globally, interpreting their implications for trading strategies and business operations, and even suggest necessary adjustments to internal policies. This is a complex area, as regulations often involve nuanced legal interpretation.
-
Market Infrastructure & Operations (High potential for automation):
- Clearing and Settlement: The entire post-trade process could be automated, with AI managing ledger updates, collateral, and payment flows.
- Exchange Operations: AI could manage the matching engines, data feeds, and underlying infrastructure of exchanges, including automated maintenance and anomaly detection.
- Network Security: AI could provide constant, self-healing cybersecurity for the entire financial network.
-
Data Acquisition & Analysis within financial markets -(Already advanced, could become fully autonomous):
- Data Sourcing and Cleaning: AI could autonomously identify, acquire, clean, and integrate vast datasets from diverse sources (e.g., satellite imagery, weather data, corporate filings, supply chain data).
- Economic Interpretation and Forecasting: Advanced AI could act as an “economic brain,” analysing global economic data, generating forecasts, and identifying interdependencies across industries and regions, without explicit human economists. It would essentially derive “meaning” from data.
Areas Where Human Input Remains Challenging for 100% AI Takeover (Even for Advanced AI):
-
Defining Purpose and Ethical Frameworks:
- “Why” Question: AI cannot spontaneously decide why a market should exist or what its ultimate goals are. These are philosophical and societal questions that require human values.
- Ethical AI: Establishing and continuously adapting ethical guidelines for AI in finance (e.g., avoiding bias, ensuring fairness, deciding on acceptable levels of risk) would still require human moral judgment.
-
Legal & Contractual Framework Creation/Evolution:
- Legislating and Adjudicating: While AI can interpret existing laws and draft contracts, the act of creating new laws or adjudicating disputes when rules are ambiguous or broken still relies on human understanding of intent, justice, and societal impact. AI cannot (yet) develop new legal principles or override existing ones based on emergent societal needs.
- Negotiation (True Negotiation): While AI can execute pre-programmed negotiation tactics, true, high-stakes, multi-faceted negotiation often involves understanding human psychology, building rapport, and finding creative solutions outside predefined parameters – skills that are difficult for AI to fully replicate.
-
Managing Systemic Confidence and Panic:
- Psychology of Markets: Markets are inherently influenced by human emotion (fear, greed). While AI can detect sentiment, it cannot reassure or panic in a human way, nor can it inherently understand the irrational components of human behavior that can drive market crashes or bubbles. In extreme crises, human leadership and communication are vital to restore confidence.
- “Last Resort” Authority: Who pushes the “off” button if the AI system goes rogue or enters an uncontrolled feedback loop? That ultimate responsibility for systemic stability would likely always remain human.
-
Innovation and Disruption (Beyond Optimization):
- Truly Novel Business Models: While AI can optimize existing businesses, radical business model innovation (e.g., creating entirely new financial products, or disrupting the very concept of money/exchange) often stems from human creativity, intuition, and a willingness to take risks that extend beyond data-driven probabilities both within and beyond financial markets.
Conclusion:
An advanced AI system could automate the vast majority of transactional, analytical, and even many strategic elements of financial markets. It could build, maintain, and optimize itself to an incredible degree, interpreting data and reacting to known patterns faster and more efficiently than any human.
However, the irreducible “human essential” activities remain those related to:
- Defining the ultimate purpose and value system.
- Creating and fundamentally evolving the underlying legal, ethical, and societal rules.
- Exercising ultimate judgment and responsibility in truly unprecedented crises where novel human responses are required.
- Injecting the fundamental human capital and economic activity that gives the market its meaning.
So, while markets could reach perhaps 95-98% AI-run in terms of execution and operational management, the “human essential” core, however small in percentage, would still be indispensable for the market’s long-term existence, purpose, and stability, particularly in navigating genuine “black swan” events or evolving societal needs.