A few years ago, astronomers found themselves burdened with a tricky problem. Having surveyed the night skies in a project known as the Sloan Digital Sky Survey, they were faced with the giant task of analysing the images of the million or so galaxies that the survey had spotted.
That’s no trivial task. Astronomers need to know things like the colour of each galaxy, whether it is elliptical or spiral and if spiral, with arms pointing in a clockwise or anticlockwise direction.
This is the kind of task that might at first glance seem easily automated, given the advances in machine learning in recent years. But the range of shapes, colours and brightness of the galaxies, their variable orientation and the amount of background detail make the task tricky for even the most powerful computers. For the moment at least, humans vastly outperform computers at this task, which is bad news for the postdocs required to spend months staring at these images.
That’s when these guys hit on a better way. Their idea was to distribute the images to members of the public and, after a short tutorial, ask them to do the classification instead. By asking lots of different people to classify the same image, researchers found that the average score was just as good as the classification given by professional astronomers.
The project is called the Galaxy Zoo and it has become hugely successful. Since 2007, hundreds of thousands of citizen scientists have supplied over 100 million classifications. The results have changed the way astronomers understand galaxy formation and evolution. The “zooniverse”, as this data is called, is now a powerful force in astronomy.
Which is why the next stage of the project looks so interesting. Instead of asking citizen scientists to classify galaxies, they are now being asked to identify supernovas, the rapid brightening of stars in their death throes.
Supernovas are so bright, they can often be spotted in distant galaxies. But the images are manifold. In the past, human scanners might have to view up to 5000 images a night during a typical observing session. Clearly a galaxy zoo approach can help.
So the team has modified the existing zooniverse software to show supernova images instead. Anybody logging in to help is shown three images of the same area of sky. The images are a “before” shot showing the region before the supernova occurred, an “after” shot showing the supernova as it brightened and an image in which the detail in the before shot has been subtracted from the after shot leaving only the supernova candidate.
The citizen scientist simply has to answer a series of questions about these images which are designed to filter out unlikely candidates.
The results are revealed today by Arfon Smith at the Univiersity of Oxford and a number of buddies. Some 13,000 people have taken part. “Citizen scientists are extremely good at identifying real supernovas,” they conclude.
That’s handy. Smith and co have used data from 48 inch Samuel Oschin telescope at the Palomar Observatory in California but there are many other surveys that could also provide data for the project.
The results from these humans could also be used to help train machines to take on the task as well. Almost certainly, the Galaxy Zoo project will one day seen as an antiquated stop gap that helped astronomers perform their work in the years before computers were powerful enough to the do the job instead.
People will one day look back on it in the same way as we look at the Victorian penchant for racing human sprinters against the first cars.
But tat’s not to downplay its value. The results will be hugely useful. Supernovas are important events. They supply the universe with heavy elements, the energy they produce triggers star formation and observations of the most distant of them provide the best evidence we have of the accelerating expansion of the universe.
But although common throughout the universe, supernovas are still poorly understood.
With a little help from the citizen scientist, that looks set to change.
Ref: arxiv.org/abs/1011.2199: Galaxy Zoo Supernovae
What to know about this autumn’s covid vaccines
New variants will pose a challenge, but early signs suggest the shots will still boost antibody responses.
DeepMind’s cofounder: Generative AI is just a phase. What’s next is interactive AI.
“This is a profound moment in the history of technology,” says Mustafa Suleyman.
Human-plus-AI solutions mitigate security threats
With the right human oversight, emerging technologies like artificial intelligence can help keep business and customer data secure
Next slide, please: A brief history of the corporate presentation
From million-dollar slide shows to Steve Jobs’s introduction of the iPhone, a bit of show business never hurt plain old business.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.