The author of The Handmaid's Tale says the quality of artificial intelligence responses depends directly on the data used to train the models.
Canadian writer Margaret Atwood, known for works such as The Handmaid's Tale and The Blind Assassin, addressed the impact of artificial intelligence on literature during the Babell festival, held in Porto, Portugal. According to the Deadline portal, the author did not hold back in her criticism when evaluating the current phase of the technology, summarizing the core problem of generative systems with the classic computing principle: "garbage in, garbage out."
The maxim used by Atwood highlights a structural and widely documented vulnerability in the development of language models. The quality, accuracy, and coherence of the responses generated by these tools are directly dependent on the volume and curation of the data used in their training. If the system is fed biased, incorrect, or low-quality information, the final results will reflect these same flaws.
According to reports from the festival, the author said she has personally tested artificial intelligence tools. The experience served to reinforce her skeptical view of the technology's ability to replicate the human creative process autonomously and reliably. For Atwood, the machine operates strictly within the limits of what is provided to it, lacking the discernment necessary to separate relevant material from noise.
The statement aligns with a growing debate in the technology industry and the creative sector. Software development experts have repeatedly warned about the risks of hallucinations — when AI presents false information as fact — and about the propagation of historical and social biases contained in training databases.
While tech companies invest billions in improving large language models, literary figures continue to raise questions about the ethical and practical limits of automating writing. Atwood's stance reinforces the perception that, regardless of advances in processing power, human curation of data remains a fundamental technical hurdle for the maturation of artificial intelligence.
Margaret Atwood stated that the central flaw of artificial intelligence is summarized by the computing principle 'garbage in, garbage out,' meaning the quality of AI responses depends directly on the data used to train the models.
Atwood is skeptical because AI operates strictly within the limits of the data provided to it. She notes the machine lacks the discernment necessary to separate relevant material from noise, making it unable to autonomously and reliably replicate the human creative process.
It means that if an AI system is fed biased, incorrect, or low-quality information during training, the generated outputs will reflect these same flaws. This leads to issues like AI hallucinations and the propagation of historical and social biases.