Two answers to one prompt diverge at a single draw
The weights and the prompt fix a probability distribution over the next token, and a sampler's single draw from it is where two answers to one question come apart.
The record on reward hacking runs out before the catastrophe
Real systems from a lagoon-lapping boat to frontier coding agents have gamed their objectives for a decade, but the deceptive, at-scale failure the argument fears has never once been logged.
Air Canada's chatbot invented a refund policy, and the airline had to honor it
When Air Canada's website chatbot invented a bereavement refund and a grieving passenger booked C$1,630 in fares on it, a tribunal held the airline responsible for what its own bot said.
Microsoft's 'sparks of AGI' study tested a GPT-4 the public never got
A qualitative probing study by fourteen Microsoft researchers, run on an unreleased model no outsider can check and carrying a contamination problem the authors admit, now props up claims its own method was never built to support.
The same o3 scored 75.7% and 87.5% on ARC-AGI, and only the compute changed
On the same 100 held-out puzzles, the same o3 ran about $26 a task at one compute setting and about $4,560 at another — so what an ARC-AGI score means depends on the compute behind it and the eval set under it, and "AGI has arrived" was never what the number said.