Will a major cosmological simulation be AI-accelerated by the end of 2027?
AITechnical AI TimelinesScienceAI Safety
Basic
23
Ṁ539
2028
68%
chance

Currently, cosmological simulations such as IllustrisTNG or Eagle are run by solving (magneto)hydrodynamical equations numerically. By the end of 2027 will at least one of the leading projects of the time (by number of citations and based on my judgment) rely critically on machine learning technology such as graph neural networks?

For this to resolve YES the main hydro/n-body simulation engine must become AI accelerated, not just the subgrid physics.

Get Ṁ1,000 play money
2 Comments
22 Holders
27 Trades
Sort by:
bought Ṁ30 YES

Seems like the sort of thing neural network potentials and stuff would be good at

Graph neural nets have been used in hydrodynamics in contexts where real time data is available (e.g. in atmospheric science) but the absence of a clear ground truth in the cosmological context may be problematic

Related questions