Introducing Your New AI Assistant
Of all the galaxies in the sky, which should you look at first?
You might have noticed that our most recent galaxies, taken by Hyper Suprime-Cam in Hawaii, have often been less detailed than our previous galaxies. This happened because we were very generous when selecting which galaxies to include. We had never classified galaxies from this instrument before and we were curious to see what you would find.
Reviewing the classifications so far, the clear majority of galaxies – as many as 90% – were voted as either “smooth” or “problem”. That suggests we’re showing you too many “blobs” and too few interesting featured galaxies.
To fix this, I’m happy to announce a new Galaxy Zoo feature that will prioritise showing you detailed and unusual galaxies.
We’re doing this with an AI algorithm we affectionately call Zoobot. Zoobot tries to classify galaxies based on what volunteers have said before for similar galaxies. If you’d like to test out Zoobot on some DECaLS galaxies, you can play with it here.
Zoobot is very good at simple classification tasks like recognising a smooth galaxy as smooth. We can therefore avoid showing you galaxies that Zoobot already considers to be very smooth. Specifically, we will now avoid showing galaxies where Zoobot is 90% confident that fewer than 2 out of 10 volunteers would click “featured”.
How will the galaxies you see change? You will see fewer totally smooth galaxies and more of everything else. Below are random examples of galaxies on Galaxy Zoo before (left) and after (right) switching on Zoobot.
Galaxies shown on Galaxy Zoo before activating Zoobot (left) and after (right). There will be fewer totally smooth and “bad zoom” images, and more featured and unusual galaxies.
Zoobot will continue to learn from you as you classify galaxies. It should get better and better at removing extremely smooth galaxies over time. If Zoobot learns that a galaxy might not be smooth after all – that is, if Zoobot changes it’s mind – the galaxy will be shown.
You might remember Zoobot from the “Enhanced” workflow we ran during GZ DECaLS. There, Zoobot tried to prioritise galaxies which, if labelled by you, would most help it learn. This worked well and helped us train the improved version of Zoobot that we’re using now. This new system is similar; we’re showing you the galaxies where volunteer labels are most useful for Zoobot and for science. We’re just using a much more straightforward rule to pick these.
Your time is precious. Galaxy Zoo volunteers can recognise and classify the detailed features of galaxies in ways that Zoobot can’t – and nor can any other algorithm. More than that, humans have a unique ability to spot things that look just a little bit weird. Volunteers talking about strange objects has led to some of our favourite discoveries, including the Voorwerpen. Using Zoobot means you will be much more likely to see more diverse galaxies and come across more weird and wonderful objects. We also have another surprise planned around these – stay tuned…
Mike, on behalf of the Galaxy Zoo team
I remember hearing years ago, on BBC R4 (6-50am) a discussion between Karen and someone from Hatfield Univ. on examining GZ images.