PDFs are fundamentally a visual layout format:

Unlike HTML or plain text, a PDF stores instructions for drawing text and graphics on a page, not the logical reading order itself. That means what you see visually isn’t what machines easily interpret under the hood. OCR (optical character recognition) can help, but it’s imperfect and often produces garbled output, especially with multi-column pages, tables, or complex formatting.

State-of-the-art models will still hallucinate:

Even advanced AI systems, when asked to extract structured data from PDFs, tend to summarize, mix footnotes with main text, or invent content that isn’t there because the AI cycles through different parsing tools and often guesses rather than understands.

Some specialized tools try to fix this:

Companies like Reducto and research models like olmOCR are training vision-augmented models specifically on PDFs, improving how headers, tables, and diagrams are detected. These approaches bring better structure and reliability than generic AI pipelines.

Volume and variety make it harder:

Billions of PDFs exist in the wild, government reports, textbooks, and legal documents, each with its own mix of text, images, diagrams, and nonstandard layouts. That diversity makes a universal parser extremely difficult to build.

Parsers are improving, but not perfect:

Newer systems trained on mixed PDF datasets can succeed in many cases, but the last bits of accuracy in the edge cases where layout tricks confuse the model remain unsolved because probabilistic models will always struggle to guarantee 100% correctness.

Pollution limits on coal plants have been rolled back:

The Trump administration’s Environmental Protection Agency (EPA) repealed Biden-era updates to the Mercury and Air Toxics Standards (MATS), reversing tighter limits on mercury and other hazardous pollutants from coal-fired power plants and reverting to older, less stringent standards.

AI data centers are contributing to rising power demand:

Across the U.S., electricity needs are increasing due to the rapid buildout of energy-intensive AI data centers alongside electric vehicle adoption and manufacturing growth, creating pressure on the grid.

Coal plants remain a major source of toxic emissions:

Coal-fired plants are among the largest sources of mercury and hazardous air pollutants, which can travel long distances, deposit in waterways, and accumulate in fish, posing serious health risks to children, pregnant women, and communities near plants.

EPA cites cost savings and grid reliability:

Officials argue the rollback will reduce compliance costs for utilities operating older coal plants and help maintain baseload power to support growing electricity demand, including from AI infrastructure.

Environmental and health advocates warn of consequences:

Critics, including public health and environmental groups, say weakening these standards puts public health at risk, undermines long-term air quality improvements, and disproportionately affects vulnerable populations already exposed to pollution.

Start a New Chat

Switch Your AI Model

Try This Prompt & Observe the Difference