2
$\begingroup$

It's not really space exploration unless space crews need to train new machine learning (ML) models for some new problems they face during their journey but I read this: https://www.technologyreview.com/2019/06/06/239031/training-a-single-ai-model-can-emit-as-much-carbon-as-five-cars-in-their-lifetimes/

Since SpaceX is planning or are launching "constellations" of satellites, as the price of launches become lower and lower, would they be able to launch satellites or spacecraft with large solar panels (ones that roll out like on the DART mission) and use solar power to train ML models? Then the trained parameters can be sent back to Earth.

Improve this question
$\endgroup$
3
  • $\begingroup$ It might be interesting to know how long it would take a satellite in space, with the best/largest solar arrays, to generate the same amount of energy as it took to get the satellite into space. $\endgroup$
    – rickhg12hs
    Dec 1, 2021 at 10:25
  • $\begingroup$ Hydrogen fuel emits water vapor but apparently there are other effects like ozone depletion and release of soot: sciencefocus.com/space/… $\endgroup$
    – mLstudent33
    Dec 1, 2021 at 23:44
  • $\begingroup$ And how much energy was required to produce the hydrogen fuel? $\endgroup$
    – rickhg12hs
    Dec 2, 2021 at 3:06

1 Answer 1

Reset to default
2
$\begingroup$

As far as I know they may already train ML models in space. It's a misconception that you need massive computing power to train a model, the computing power you need is dependent on the size of the data set, the number of features you want to run the model on and the type of model and algorithm you want to use. You can train a model with very little power if you aren't doing anything that complex, I personally trained ML models using a Raspberry PI. Where you need massive power is when you are running extremely complex calculations like weather forecasting, and when working on huge data sets like large populations of people.

Improve this answer
$\endgroup$
3
  • $\begingroup$ What kind of models did you train on a Rasberry PI? $\endgroup$
    – mLstudent33
    Dec 1, 2021 at 23:36
  • $\begingroup$ I did some text work around sentiment analysis and image recognition training @mLstudent33. $\endgroup$
    – GdD
    Dec 2, 2021 at 20:41
  • $\begingroup$ Good to know you can pull that off on a Rasberry PI. $\endgroup$
    – mLstudent33
    Dec 3, 2021 at 6:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.