The AI Backlash and a Looming Bubble

Public frustration with AI-generated media is intensifying, from vandalized ad campaigns to artists condemning unauthorized cloning, amid critiques that AI is inauthentic and exploitative. At the same time, investment in AI infrastructure is skyrocketing, backed by tech giants and U.S. policy, despite weak signs of real demand and mounting legal risks. Analysts warn spending is outrunning plausible returns, pointing to a looming bubble whose impact could ripple across the broader economy.

Key Points

• Public sentiment toward generative AI is souring as people mock and deface AI-driven ads and content, criticizing a loss of authenticity and value.
• High-profile artists and critics decry cloning and synthetic media ethics, while experts argue the technology is oversold and built on exploitative data and labor.
• Despite skepticism, investment is surging—hundreds of billions in data centers and initiatives like the $500B Stargate project—driven by Big Tech and U.S. policy.
• Economists warn spending is outpacing plausible returns, with circular funding and limited real customer demand raising bubble risks.
• Legal and IP challenges (e.g., NYT vs. OpenAI) and style imitation controversies deepen uncertainty as profitability remains elusive.

Sentiment

The community is predominantly skeptical, driven mainly by the debunked 'world's first' claim and the underwhelming quality of the output. There is mild curiosity about the technical pipeline and some sympathy for the democratization argument, but the overall tone leans critical with a hint of gentle mockery.

In Agreement

• The diffusion-to-TTF pipeline (generate raster glyphs, binarize, vectorize with Potrace, compile to TTF) is a genuinely creative technical approach worth documenting.
• Custom font design has historically been prohibitively expensive (the article's $2,000/character example resonated), and AI-based tools could reduce costs and democratize access.
• The iterative prompting approach with typographic grids to improve normalization and alignment is a clever workaround for the consistency challenge in AI-generated typography.
• The writing style of the article was appreciated by some as refreshingly energetic and honest about the iterative, imperfect process.

Opposed

• The 'world's first AI-generated font' claim is false — multiple prior implementations exist, including Tom7's 2021 AI font, a June 2024 SD 1.5 pipeline on GitHub, gwern's dropcap workflow, and years of academic research.
• The actual font quality is poor and the article does not demonstrate the font in real-world use as running text, which is the most meaningful test for a typeface.
• The vision of every individual or Substack publisher having a custom AI font was met with horror rather than enthusiasm by much of the community.
• Even by the article's own admission, normalization, baseline consistency, and metric management remain unsolved challenges that leave the font unusable for professional purposes.