Astrometry.net and AI

[László Heller]

Hi Dustin and folks!

Could be possible that at some point the astrometry.net engine will be (fully or partially) replaced by artificial intelligence (ChatGPT & co) ?

BR,

Ladsislav

[Martin Sadler]

I would say that the astrometry app is already artificial intelligence!

--
You received this message because you are subscribed to the Google Groups "astrometry" group.
To unsubscribe from this group and stop receiving emails from it, send an email to astrometry+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/astrometry/8212cf9b-151b-4942-bf65-4c22da1234efn%40googlegroups.com.

[Dustin Lang]

The "geometric hashing" idea that we use was originally developed by people thinking about the human visual system and our ability to see part of an object and think about the whole object.  Like "oh, that shape reminds me of..."

[Duncan Carlsmith]

I'm also interested. The use of scaled asterisms is quite clever but does this scale to GAIA and beyond? Soon we will have gigapixel imagery of the entire galaxy not just 3/4. Newbie me, I'm not sure how to implement this but am psyched to try. Of course, paring the problem down with known information helps. Navigation by the stars through our own solar system - how is that done in practice?

[Dustin Lang]

Hi,

The nova.astrometry.net web service uses index files built from Gaia.

On celestial navigation -- within the Solar system, the distances relative to the stars are pretty small -- compare the distance to the Voyager spacecraft to the distance of the nearest star and it's a tiny fraction.  So most stars don't move much.  They did some experiments using the New Horizons spacecraft to measure some stellar distances by parallax.

cheers,

dustin

To view this discussion on the web visit https://groups.google.com/d/msgid/astrometry/f3ae5404-60fb-417a-85c7-c86979842f12n%40googlegroups.com.

[paul.l...@gmail.com]

Dustin wrote: "So most stars don't move much."

Yes and no.

While stellar parallax is rather small, less than 1 arcsec for all known stars, each of aberration and nutation can reach up to 20 arcsec or so. The latter pair are of no interest for solving to a specific epoch such as 2000.0 as used in the astronomy.net tables. All these effects change the observed position cyclically over the course of the year.

The killer, though, is proper motion which is cumulative (positions change in a specific direction and its magnitude always increases); a motion of 1 arcsec per annum is not too unusual and > 0.3 arcsec per decade is rather common. The extreme example is Barnard's star which moves by about 9 arcsec every year. I have an animated GIF made from images I took each summer which shows it moving against the background stars. It is really quite remarkable.

Now the above might sound rather theoretical and inconsequential, but I have been badly bitten by proper motion. AAVSO provides positions of variable and comparison stars. All their positions are precessed to J2000 but the underlying astrometry often dates back to the 1950s.  On numerous occasions, matching AAVSO data with Gaia-EDR3 values has required human intervention because the stars have moved by several arcsec in the interim.

Paul