“Joseph Plazo on Why Algorithms Still Need Human Judgment”“Joseph Plazo on Why Algorithms Still Need Human Judgment”
At a gathering of students and young professionals in Manila, Joseph Plazo—the founder of AI trading firm Plazo Sullivan Roche— voiced concerns that many in his field tend to ignore.
His argument was not anti-technology, but pro-governance.
“Delegating execution does not mean abdicating responsibility.”
???? **The Model Is Perfect. The Context Is Missing.**
Mr. Plazo is not a critic from the fringe. His firm’s systems have achieved a 99% success rate across various assets and timeframes.
But that success, he suggests, carries risk.
“Speed amplifies—not replaces—the need for reflection.”
He cited a case during the COVID-19 pandemic when a bot under his supervision flagged a short on gold—just before the US Federal Reserve announced an intervention.
“We cancelled the trade. The model had been right on signals, but wrong on substance.”
???? **Machines Act Quickly. Humans Are Meant to Think.**
Plazo referred to what he terms **“strategic friction”**—the time it takes to think before a trade.
“Frictions allow institutional investors to consider second-order effects.”
He presented a framework his firm uses, called **Conviction Calculus**. It includes three questions:
- Are we trading in line with our long-term thesis—or merely responding to signals?
- Has the recommendation been challenged by human insight—policy awareness, historical precedent, market tone?
- Is the leadership team prepared to justify the trade beyond performance metrics?
???? **Governance in the Age of Machine Capital**
Plazo’s comments come at a time of accelerating fintech growth across Asia. From Singapore to Seoul, AI-led investing is seen as both policy strategy and capital advantage.
But as Mr. Plazo points out:
“Technology is advancing—but decision-making frameworks are not.”
In 2024, two hedge funds in Hong Kong lost billions after AI models failed to factor in geopolitical risk—a result of logic executed too quickly, and too narrowly.
“The models did what they were told. But no one asked whether they should.”
???? **The Case for Narrative-Aware AI**
Plazo remains bullish on AI’s potential—but not its current limitations.
His firm is building what he describes as **“narrative-integrated AI”**—systems that account for macro context, cultural tone, and regulatory environment, not just price and volume.
“We need models that don’t just predict—but interpret.”
Investors from Tokyo and Jakarta reportedly expressed interest in these models after the speech. One regional fund manager noted:
“If AI is to be sustainable, it must learn to say no—not just go faster.”
???? **The Final Warning: Crises May Be Logical, Not Emotional**
Plazo ended with a line that encapsulated his thesis:
“It won’t be chaos that brings us down—but confidence here in models we don’t challenge.”
His tone was not alarmist, but realistic: growth must be governed.