Yuval Noah Harari: NEXUS in TivoliVredenburg Utrecht, The Netherlands
"ChatGPT lied," Yuval Noah Harari declared during our recent conversation at TivoliVredenburg in Utrecht, where he presented the Dutch translation of his latest book "Nexus." His statement, while compelling, reveals a deeper philosophical confusion about the nature of deception and truth.
The accusation of "lying" presupposes consciousness, intent, and an understanding of truth - qualities that large language models fundamentally lack. These systems don't make truth claims in any meaningful sense; they generate probabilistic responses based on patterns in their training data. When a human lies, they consciously choose to deceive, knowing the difference between truth and falsehood. An AI system can produce incorrect information, but calling this a "lie" anthropomorphizes a statistical process.
During our discussion, I challenged Harari's framing. The real issue isn't that AI systems lie or tell the truth, but rather that they operate in a space entirely outside the traditional binary of truth and falsehood. They are sophisticated pattern matchers generating outputs that can appear convincing while being disconnected from any genuine understanding or truth-seeking behavior.