Comment by Kye

Comment by Kye 2 days ago

4 replies

Machine learning (AI) is used everywhere in astronomy. That's how they made the black hole image. Don't confuse the broader 60+ year old world of ML with transformers and diffusion models.

cma 2 days ago

Not sure if there was an update/response to this but:

     1st image of our Milky Way's black hole may be inaccurate, scientists say
https://www.space.com/the-universe/black-holes/1st-image-of-...
vlovich123 2 days ago

Based on the paper I linked, it seems like a straight up classical sampling and clustering with baysian hyperparameter tuning. This is “everything is now AI” slop that’s infected all grants, academic and private industry fundraising. There’s no neural net or LLM involved.

  • refulgentis 2 days ago

    That's machine learning.

    Clustering alone is machine learning and has been taught as such to innumerable people.

    I have deep feelings about this, someone in management taking exactly one Kaggle course managed to wield this knowledge to great damage.

    But it is machine learning.

    Additionally, it goes far beyond clustering: the article you linked describes training an image recognition model, which also seems to be heavily stressed in the article linked on HN.