Q&A: How AI came to model the “messy” human mind
Assistant Professor Benjamin Lindquist studies how engineers deliberately injected randomness into early AI, giving rise to deep-learning models that sometimes hallucinate.
Early computer engineers believed injecting randomness and irrationality into AI systems was essential to mimic the creative thinking of humans, contends Benjamin Lindquist, an MIT researcher. In a new paper, Lindquist traces the origins of AI hallucinations to this early research, suggesting tradeoffs between creativity and predictability when building machines that mimic human behavior.
The belief that chance plays an important part in discovery and creativity dates back to mid-20th century science and art. John Pierce, a prominent electrical engineer at Bell Labs, argued that removing order from art through math could add bursts of the unexpected, leading to insight. Psychologists believed artificial neural networks could model the brain and reveal the "messy" aspects of human thought, unlike "black box" systems.
Early AI systems were viewed differently than today's ChatGPT or Claude. Advocates valued them for providing unexpected, creative answers, while modern systems are more focused on delivering correct, expected answers. This shift can be attributed to the operational differences between early laboratory-based systems and today's widely used AI, which must consider legal and monetary consequences.
Neural networks, initially seen as valuable for generating unexpected outputs, eventually became the dominant paradigm in AI due to advancements in data availability and computational power. Around 2012, neural networks outperformed statistical machine learning, leading to their growth and dominance in AI development. Studying AI history can provide valuable lessons, cautioning against suppressing creativity in favor of predictability.
Written by urgent.news from MIT News Research's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
