Large Language Models Will Never Be Intelligent, Expert Says

Prominent critics and recent academic work argue that large language models (LLMs) emulate language without replicating human thought, undermining claims that current LLM architectures can produce artificial general intelligence (AGI). Citing neuroscience, case studies of language loss, and a mathematical analysis of creative limits, the critique calls into question multi‑billion dollar scaling strategies—heavy GPU capex and data‑center buildouts—while flagging environmental and innovation risks to companies betting on LLM‑driven AGI.

Analysis

Market structure: If LLMs hit a practical ceiling as argued, capital intensity in pure LLM scaling could decelerate within 12–24 months, favoring hardware/service providers (NVIDIA, AMZN, MSFT) that can repurpose capacity vs. software firms (including META) whose AI pitch supports valuation multiples. Pricing power shifts to suppliers of diversified compute and edge/robotics IP; vendors locked into LLM-only roadmaps face margin compression as clients demand multi-modal/world models. Short-term demand for GPUs and cloud remains strong, but supply/demand may flip from seller’s to buyer’s market if capex growth slows from recent double-digit rates to low single digits over 2025–26.

Key Decisions for Investors

• Establish a 2–3% long position in NVDA (NVIDIA) using a 6–12 month horizon; layer on additional 1–2% if price retraces >10% from today or implied vol declines — target hold unless NVDA reports QoQ revenue contraction >15% or gross margin collapse.
• Trim/evaluate reducing META exposure by 30–50% over the next 4–8 weeks ahead of strategy updates; hedge remaining exposure with a 3-month put spread (buy 1x 10% OTM put, sell 1x 20% OTM put) sized to cover 1–2% portfolio risk.
• Implement a pair trade: long NVDA / short META equal notional (1:1) for 3–6 months to capture hardware demand vs. pivot execution risk; size to 1–2% net portfolio delta and rebalance on earnings-driven moves >8% intraday.
• Rotate 2–4% of portfolio from small-cap AI services into MSFT, AMZN, GOOGL and EQIX over 3 months (each 0.5–1% increments); exit or re-evaluate if any single name reports FY guidance beat that materially boosts AI monetization (>10% forward rev guidance upgrade).
• Use options for asymmetric risk: buy 6-month NVDA call spreads (debit) to cap cost and buy 1–3 month META put spreads into next two earnings windows; limit total options premium to no more than 1–3% of portfolio exposure to these themes.