Margaret Atwood on AI's Core Flaw: It Can't Think, Only Regurgitate

Margaret Atwood, the celebrated author of The Handmaid's Tale and The Blind Assassin, shared pointed criticism of artificial intelligence during an appearance at the Babell Literary and Cultural Festival in Porto, Portugal. She offered a bracing take on why AI systems so often fall short of their hype.

A Disappointing Trial with Claude#

Atwood recounted her single experiment with an AI tool. She tried Anthropic's Claude chatbot once—and once was enough. She had posed a straightforward question: details about the British detective series Father Brown. The response, she found, was simply wrong. Rather than delivering accurate information, the AI had produced what amounted to an educated guess built on flawed reasoning.

"Claude gave me the wrong answer, or it lied," she explained, then clarified the underlying issue. The system couldn't help but mishandle the task because it lacks conscious understanding. Instead, it was working from patterns in its training material—including television reviews that strategically avoid revealing plot details. Starved of complete information, it fabricated a plausible-sounding but incorrect answer.

The Real Problem: Garbage In, Garbage Out#

Atwood distilled the technological problem to a simple principle. AI systems inherit the limitations of their source material. They're trained on internet text, published criticism, and data pulled from countless websites—much of it potentially biased, incomplete, or outdated. Feed imprecise or misleading information into an AI model, and it will reliably produce confused output. No amount of processing power overcomes bad inputs.

She was equally direct about the people deploying these tools. Those seeking shortcuts through AI were essentially "opportunists" taking the easy route—and she wasn't gentle about their prospects. Even those using AI for legitimate business purposes cannot simply trust its output without verification. The system makes mistakes routinely, demanding human oversight to catch errors.

The real risk, Atwood suggested, isn't that AI will replace human judgment—it's that people will convince themselves it already has.