Why AI is far less useful than they say | Ilyaz Nasrullah | Money & Macro Talks

The Gist

Artificial intelligence tools and large language models are vastly overhyped and fail to deliver on productivity and intelligence promises because they lack true understanding and rely on flawed, automated processes.

Quick Overview

Artificial intelligence models are fundamentally limited by a lack of real understanding, meaning they operate as powerful automated tools rather than revolutionary cognitive entities. Host Dr. Joeri Schasfoort and computer scientist Ilyaz Nasrullah dissect the seven major myths of artificial intelligence, exposing why adoption does not automatically equal productivity and why hype cycles risk distorting economic reality.

Key Points: Ilyaz Nasrullah published a book in Dutch titled This Is AI, focusing on the seven false promises of artificial intelligence. A Harvard study by Fiona Chen tracking seven hundred companies and seven hundred thousand software workers between January 2021 and March 2026 found that coding agents increased individual task flags by twenty-three percent, but produced a fifty percent increase in review time and a thirty-five percent increase in review comments. A Jellyfish data set analyzed in the research revealed zero organizational productivity gains despite widespread individual software tool adoption. Google introduced the Facts Suite benchmark in December 2025 to evaluate the factual accuracy and grounding scores of automated models. The 1990 chess match where IBM Deep Blue defeated Garry Kasparov demonstrated computational brute force rather than true human cognitive intelligence.

Context: Dr. Joeri Schasfoort hosts Money and Macro Talks to explore macroeconomic realities behind major technological trends. In this episode, he sits down with computer scientist, columnist, and AI advisor Ilyaz Nasrullah to examine the economic and technical boundaries of large language models.

Detailed Analysis

The conversation deconstructs the grand promises made by tech executives like Sam Altman and Dario Amodei, comparing their utopian narratives to economic reality. While individual developers experience short-term efficiency boosts when using coding agents, organizational metrics reveal that code quality drops and review overhead surges. Nasrullah argues that artificial intelligence systems use brute force computation rather than true comprehension, creating a dangerous illusion of intelligence that mirrors past market bubbles.