📚 How to solve these questions: 12 tips for all 6 Interactive Reading question types, with a worked example for each.
For most of its history astronomy was limited by what a human eye could hold at a telescope, and progress meant building a 1 mirror. Modern astronomy is limited by something else entirely. The instruments now produce more data every night than any team can examine, and the discovery problem has moved from collecting observations to finding the interesting ones inside a flood of ordinary ones. A single modern survey telescope photographs the whole visible sky every few 2 and generates tens of terabytes in the process. Almost all of it is unremarkable. Somewhere in it may be a supernova that started yesterday, or an asteroid on a path worth checking, and both are only useful if they are found 3 enough for another instrument to look. The response has been automation. Software compares each new image against a reference, flags whatever has changed, and 4 the results by how likely they are to matter. This works well and it has a known 5. A system trained on what has been seen before is very good at recognizing familiar events and systematically poor at noticing something with no precedent. That is why several projects deliberately route a fraction of the unusual detections to human volunteers rather than filtering them out. The reasoning is not sentimental. If the genuinely new is by definition unlike the training data, then some part of the pipeline has to be capable of finding it interesting for reasons the software cannot state.