>>723964929
From what I understand, machine learning algos are fairly useful in astronomical research because they can quickly and reliably accomplish certain identification/classification tasks that are tedious for humans (typically things like comparing two plates flipping back and forth trying to play 'spot the difference' to identify new supernovae or measure the varying brightness of a given star).

I think transformer/machine learning tech can be useful when it's trained for a specific task and put to work on that task. I don't think that the whole "treat a chatbot as a general intelligence because mimicking language lets it fake expertise" has actually panned out for anyone ever except maybe middle schoolers cheating on homework assignments.