What is going on with AI?

2026-04-20 · Source: David Shapiro · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure · Depth: Intermediate, extended

Summary

The current AI buildout, particularly in data centers, represents the second-largest privately funded mega-project in history, surpassed only by the Marshall Plan in terms of GDP percentage. This massive investment, often misconstrued as an "AI bubble," creates durable assets like data centers, which last 50+ years, and GPUs, which depreciate but retain value and offer tax benefits. The article critiques academic perspectives, exemplified by writers like Cal Newport, who often rely on outdated studies and lack practical industry experience. These studies frequently use older AI models (e.g., ChatGPT 3.5, Qwen 2) and flawed methodologies, leading to incorrect conclusions about AI's capabilities and productivity impact. The rapid pace of AI development outstrips academia's slow publication cycle, rendering many published findings irrelevant by the time they appear, creating a significant knowledge gap between academic research and real-world industry application.

Key takeaway

For Directors of AI/ML or investors assessing the AI landscape, recognize that academic studies on AI's limitations are often based on outdated models and methodologies, lagging current capabilities by years. Your decisions should prioritize real-time industry insights and practical applications over slow-moving academic publications, as the rapid pace of AI innovation quickly renders such research irrelevant. Be wary of claims that dismiss AI's productivity gains without considering actual power user experiences or the latest frontier models.

Key insights

Academic research on rapidly evolving AI often lags industry reality by 2-3 years due to structural publication delays and outdated models.

Principles

• Private mega-projects create durable assets.
• Academic distance can hinder practical understanding.
• Anecdotal evidence can be crucial in dynamic fields.

Method

When evaluating AI claims, scrutinize the study's methodology, the AI models used, and the authors' industry experience, as academic publication cycles often render findings obsolete.

In practice

• Prioritize real-time industry insights over slow-moving academic papers.
• Question studies using AI models 2-3 years old.
• Advocate for AI use in educational programs.

Topics

• AI Data Center Investment
• Academic Bias in AI Research
• AI Model Obsolescence
• AI Productivity Enhancement
• Durable Capital Assets

Best for: Director of AI/ML, Investor, Consultant

Related on AIssential

• AI Is Slowing Down — The AI industry's financial model is unsustainable, requiring trillions in revenue by 2030 that current demand cannot support.
• I Can’t Keep Up With AI. That’s Not the Problem I Thought It Was. — The rapid pace of AI tool development renders traditional deep skill mastery obsolete for designers.
• Measuring the Self-Reported Impact of Early-2026 AI on Technical Worker Productivity — Self-reported AI productivity gains, though often overestimated, are best measured by "value" and complement other assessment methods.
• Are modern ML PhDs becoming too incremental, or is this just what research looks like now? [D] — ML PhDs often prioritize incremental benchmark improvements due to academic and industry incentives.
• Meta Keeps Delaying the Release of Its New AI Model to Developers — Meta's repeated AI model release delays challenge its monetization strategy for frontier AI investments.
• The Calm Before the AGI Storm — Major AI labs are aggressively positioning for a future of accelerating AI capabilities and market dominance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by David Shapiro.