Interesting object

https://www.legacysurvey.org//viewer/?ra=5.6387&dec=-31.4495&layer=ls-dr9&zoom=16&spectra&const

Have you also tried searching in VizieR for known lens (candidates) with your coordinates?

Sugohi in particular

You’ll find many many of them already found by the algo’s

For as far as my limited knowledge goes I’d wager the fields of gravitational lenses & supernovae / transients are already completely dominated by A. I. / algorithms/ robots / neural nets etc.

More fields will follow and right now it seems to me the window of opportunity for citizen science discoveries that started with digital sky surveys and the likes will slowly be shut down again by domination of these types of software

Some hope for a combined effort of human & AI, but I doubt this will be the case for the long term

Heck I even saw a paper on arxiv about an algorithm specifically trained to find the (astro) odd things and outliers, and was good at its job!!

Ok storytime over, hope this helps :joy:

Feel like you’re right. I know one of a research team that worked on the JWST’s PSF and they are all now throwing AI and machine learning at finding lensing candidates.

On the domination of citizen science, I feel you might be right but there are also definitely some areas where it will take a while. We met spot stuff fairly easily but some of the machine learning models I have seen are comical at spotting certain types of (to us) trivial patterns. Unless you have (computing) time I reckon it’ll take a while.

1 Like

Agreed, it will take awhile. And the success of The Backyard Worlds team really is something no one could have seen coming, really amazing.

But an advanced neural net trained to find brown dwarfs with inclusion of the odd thing / outliers? Everything that is found now will add to the training data of these enitities. It might take a few more years but I expect such software to easily replicate BYW succes and beyond, just like in the g-lens & transient fields atm. And these will be used on data before it gets public! How to beat that?!

Ofcourse humans will be needed after initial data selection, but not citizen scientists I’m afraid…

True. it won’t be a slow or linear takeover it will be a complete explosion as eventually enough data for training is available and continually generated that it causes an explosion in the level of these entities. I reckon 3-6 years max. Let’s face it - all these citizen science projects don’t have very high levels of detecting stuff either

1 Like

Hmm definitely don’t agree with your last statement. Looking at all the objects / phenomena found by Zooniverse astro projects, or all publications + references to those papers from all Zooniverse astronomical projects it has been a massive succes. Ofcourse initially Galaxy Zoo started out as contribution by classification, and even that is a success.

But also ongoing finds from BYW, and yet-to-be-published new Voorwerpjes (38 new ones and counting), I’d conclude that detection rate for certain objects is still very high.

And keeping in mind the upcoming algo boom NOW is the time to find stuff :slight_smile:

1 Like

That’s fair enough but I think a good machine learning model (emphasis on good - we aren’t there yet) will not miss a lot of the things that I reckon have been missed by citizen science. It has been very successful but just look at that AI analysis of radio signals that came out Monday 30th (I think). Professional astronomers missed these 4 signals from outer space (ages ago) that an AI flagged as possible extraterrestrial origin and deserving a follow up. If academics can miss some data on a relatively small data set I am sure there are some intricacies that are missed by citizen science. I guess my point is that it is very good for the majority of stuff but it misses intricacies and small details imo.

Looking forward to the “algo boom”. I love astrobiology (aka I never grew up and want to find aliens) and the paper in which really old data had been sorted through to find new stuff for the SETI filled me with joy. I reckon within 5 or so years these algorithms will not only be able to find new stuff from new data but also look through the huge amounts of data we have and spot things we never spotted

1 Like