Back to the Future: Investing With Human Intelligence in an AI-Driven World

The article argues that AI should not be used as a substitute for investing strategy, citing the 1998 LTCM collapse and its 25-to-1 leverage as a warning about overreliance on models. It promotes Human Intelligence, portfolio review, and risk management over algorithmic decision-making, while lightly promoting InvestingPro as a data tool. The piece is commentary-driven and contains no company-specific earnings or macro policy catalyst, so direct market impact appears limited.

Analysis

The real market implication isn’t “AI is overhyped,” it’s that AI-generated stock-picking content is becoming a retail flow amplifier. When a theme is packaged as certainty, it tends to pull in short-duration capital chasing momentum rather than underwriting fundamentals, which can extend moves in the near term but also make reversals more violent once the marginal buyer disappears. The beneficiaries are the platforms, newsletters, and brokerage ecosystems monetizing engagement; the losers are late entrants buying at the top of an attention cycle.

Second-order, this kind of messaging tends to favor lower-quality, high-beta names most exposed to sentiment and positioning rather than durable cash-flow compounders. If the “AI-picked winners” basket is crowded, the cleanest expression is not to fight AI broadly, but to identify the most crowded long leg and fade it into strength while owning the underlying infrastructure winners that monetize actual AI spend. The setup is more about flow and narrative dispersion than about the absolute direction of the AI theme.

The biggest contrarian miss is that investors may confuse personalization with forecasting edge. Portfolio construction failures usually come from leverage, concentration, and liquidity mismatch, not from the absence of a good stock-picker; that means the next dislocation will likely come from crowded beta exposure, not from a failure of AI sentiment itself. Time horizon matters: sentiment can stay elevated for weeks, but the unwind risk rises over 1-3 months if earnings do not validate the move or if broader risk appetite rolls over.

For us, the actionable edge is to separate “AI infrastructure” from “AI attention.” The former should continue to benefit from capex, cloud, and model deployment, while the latter is vulnerable to mean reversion once performance-chasing exhausts itself. If this narrative is already producing 15%+ monthly gains in multiple names, that is often a sign to tighten risk, not add it.

Key Decisions for Investors

• Short the most crowded AI/sentiment-driven outperformers into strength over the next 2-4 weeks; use a defined-risk put spread or call spread sale on names with weak profitability and high retail ownership. Target 10-15% downside with 2:1+ reward/risk if momentum fades.
• Long AI infrastructure beneficiaries such as NVDA, MSFT, or AMZN on 3-6 month timeframes, funded by shorts in speculative AI-adjacent names. Risk/reward favors durable capex capture over narrative names; stop if hyperscaler capex slows materially.
• Pair trade: long semis/compute infrastructure (SMH or NVDA) vs short a basket of high-beta, non-profitable AI “application” names. The thesis is that monetization will concentrate at the picks-and-shovels layer; aim for 5-8% relative outperformance over 1-2 quarters.
• If retail sentiment indicators spike further, add a tactical volatility hedge via QQQ puts 1-2 months out. This is a portfolio hedge against an unwind in crowded AI momentum rather than a directional macro short.