John Carmack on AI in game programming

AI as Power Tool vs Copy Machine

• One side agrees with the “power tool” framing: AI is seen as an extension of prior automation (cameras vs painters, Photoshop, game engines), letting skilled creators do more, and enabling small teams.
• The opposing view calls current generative systems “copy machines,” not true creative tools—built on large-scale copying of others’ work and sold as something far more magical and replacement-ready than they are.

Content Flooding, Quality, and Search

• Many worry that dramatically lowering barriers leads to a flood of low‑quality “slop,” making it harder for good work to be discovered and hollowing out the “middle class” of games.
• Others argue that 90% of everything has always been bad; the problem is and always was discovery and curation, not production.
• Algorithmic curation is criticized for optimizing for “sellable” or engagement‑maximizing content, not quality, which further drowns out good work.

Comparisons to Past Media Shifts

• Printing press, YouTube, and music/film democratization are invoked both as precedents (“people complained then too”) and as warnings (enabled propaganda, content farms, and IP-exploitation franchises).
• Some suggest we might have been better off with stronger gatekeeping/tastemakers, given today’s flood of low‑effort content. Others value niche and educational work that would never pass traditional gates.

Ethics, IP, and “Theft”

• Strong thread around models being trained on copyrighted works without consent; many call this straightforward infringement and “theft,” especially when the same workers are then displaced.
• Counterarguments liken training to human inspiration and learning; critics respond that scale, exactness of reproduction, and lack of compensation make this qualitatively different.

Jobs, Politics, and Economic Structure

• Anxiety about AI replacing significant numbers of creative and programming jobs, in a system where livelihood is tightly tied to productivity and profit.
• Distinction drawn between technological progress and political/economic fallout; the real “illness” is seen as wealth concentration and corporate consolidation.
• In games specifically, some argue AI might ease ballooning AAA costs; others say major expenses are marketing and risk-averse business models, which AI won’t fix.

AI in Coding and Learning

• Mixed views on AI for professional code: seen as tech‑debt‑prone and risky for critical systems, but potentially useful for learning, “vibe coding,” and hobby projects.