Author Margaret Atwood warned at the Babell Literary and Cultural Festival in Porto, Portugal, that artificial intelligence models remain unreliable due to their reliance on “garbage in, garbage out” data collection. Atwood, speaking during a career Q&A, noted that LLMs often make mistakes—citing a personal experience where Anthropic’s Claude failed to provide accurate plot details—and cautioned that the technology’s tendency to scrape existing, previously published works makes it a risky tool for factual reliance.
Why AI reliability remains a barrier
Large language models (LLMs) like Claude operate by sampling vast datasets, a process that inherently limits their accuracy, according to Atwood. When she tested the model for a spoiler regarding the series Father Brown, the AI provided a false answer. Atwood explained that because the system had processed online criticism—which rarely reveals show endings—it was misled by the things it had read about the show.
Margaret Atwood noted that even when AI tools are used for business, people have to check it because it makes mistakes.
The intersection of censorship and literature
Book bans are often framed by authorities as moral corrections, but Atwood views them as a good sales gimmick. She observed that labeling a book as “too hot to read” frequently triggers increased public interest. According to PEN America, Atwood’s The Handmaid’s Tale was included on a list of the most-banned books in school districts across the United States. While she resides in Canada, where she credits a commitment to freedom of speech and thought, she expressed concern that similar censorship efforts in the U.S. represent an attempt to stifle political dissent, which she describes as the lead-up to a dictatorship.
How media consolidation shapes public discourse
Atwood highlighted a shift in how political information reaches the public, noting that billionaire pro-Trumpers are buying up legacy television stations and newspapers. She pointed to a counter movement that is bringing the establishment of new media and outlets with large distribution online.
Pro tips for navigating AI information
- Verify, don’t trust: Never treat an AI-generated answer as a primary source for critical facts.
- Cross-reference: Always check AI outputs against established, human-written journalism or peer-reviewed databases.
- Understand the source: Remember that LLMs are predictive engines, not search engines; they aim to mimic human language, not necessarily to provide objective truth.
Frequently asked questions
- Is AI capable of replacing human writers?
- According to Atwood, the technology, and the way large language models are developed through scraping through previously published works, simply isn’t reliable enough for humans to depend on.
- Why are books like The Handmaid’s Tale still being banned?
- PEN America reports that these books are often targeted for their depictions of power, gender, and politics, though Atwood argues that such bans often backfire by drawing more readers to the material.
- What is the biggest risk of AI in business?
- The primary risk is the “garbage in, garbage out” phenomenon, where businesses might integrate incorrect data into their workflows, leading to mistakes.
What are your thoughts on the role of AI in creative writing and information gathering? Share your perspective in the comments below or subscribe to our newsletter for more updates on the intersection of technology and culture.
Related reading