factual

AI Used For Anomaly Detection In Big Physics Data

AI in the sense of neural networks and deep learning is already used routinely in science: at the Large Hadron Collider so much collision data is produced that only one in a million collisions can be recorded, so neural nets must rapidly decide which events are interesting and then search the resulting database for statistical anomalies—but this is constrained, well-defined problem-solving akin to AlphaGo, whereas the real goal of an AI that could invent a theory of gravity better than Einstein's remains beyond current ability.

factualpending

Speaker

Sean Carroll

Evidence Quote

“we can actually only record to tape one out of a million collisions because we just can't record core the data fast enough”

Source

124 - In Search Of Reality— Making Sense with Sam Harris
Created: 6/14/2026, 2:37:20 AM

Your notes

Loading notes...