CONCEPT
Simulation vs. Possession of Intelligence
Toby Walsh's central distinction: between machines that produce outputs resembling the products of a reasoning mind and machines that actually reason—a gap that current AI, however impressive, does not cross and that grounds every ethical and political demand he makes.
The distinction between simulating intelligence and possessing it is the load-bearing beam of
Toby Walsh's entire body of work. It underlies his skepticism of
singularity predictions, his ethics of accountability for
large language models, his campaign against autonomous weapons, and his proposal for a Turing red flag law. The distinction is simple to state and slippery to maintain. A machine that simulates intelligence produces outputs that look, on the surface, like the outputs of a reasoning mind: fluent text, accurate classifications, plausible answers to questions. A machine that possesses intelligence actually understands what it is doing—can explain its reasoning, recognize its errors, generalize robustly to situations it has not encountered, and be held meaningfully accountable for its outputs. Walsh's career-long verdict is that the present systems, however dazzling, live entirely on the simulation side of this divide. They are extraordinary engines for predicting which word, classification, or output fits best given