Robots at the telescope

Photo: mdrs.marssociety.org/

Mars Society Observatory Director Peter Detterline with a robotic telescope.

A 1950s picture of Edwin Hubble shows him using the 200-inch (five metre) optical telescope at Mount Palomar.

Like all large telescopes, the instrument uses a concave mirror rather than a lens as the light collector. Mirrors are easier to support and can be at the base of the telescope, so they can be much larger than lenses, while presenting fewer design and construction problems.

The downside is that the light collected by a concave mirror is brought to a focus in front of it, so that is where the instruments are located, in a special, unheated enclosure.

Back in the 1950s, telescopes and the instruments used on them were manually controlled, so the astronomer had to be in that enclosure too.

For the several hours or more needed to collect enough light to give a usable image or spectrum, the astronomer had to manually make sure the telescope remained centred on the object of interest.

The outcome of his or her efforts was usually just one or two images or spectra on glass slides, or even, for some projects, just a few measurements recorded in a notebook.

When computers first found their way into observatories, they were put to work controlling the telescopes, reducing the need for astronomers to spend their nights freezing while observing. The next step was when computers took over the task of recording the data.

Astronomers were still at the telescopes, overseeing observations to ensure that nothing went wrong, but basically a successful observing session would be spent sitting in a warm telescope control room reading ancient magazines, looking at computer displays and trying to stay awake.

Been there, done that!

We are now in the midst of a new revolution, with computers and digital processing moving right into the instruments as well as the telescopes.

Modern facilities are almost entirely digital. The power of modern computers and signal processors is what makes possible the new generation of telescopes, such as the CHIME radio telescope at our observatory.

This also means modern telescopes can do far more than the instruments of a decade or two ago. For example, instead of imaging a single object we can image a large patch of sky, including lots of interesting objects. The result of this is that the instrument outputs a torrent of data.

A few years ago, this was no problem. Digital data systems could record it and after the astronomer had searched through it for what he or she wanted, that raw data would be put into data centres, like the Canadian Astronomy Data Centre, for other astronomers to search through, looking for other things.

This is why we like to record the raw data. Processing to find something could erase something else, possibly an important discovery.

Instruments like CHIME produce a huge amount of data. Projects like the Square Kilometre Array radio telescope will produce not a torrent of data, but a tsunami, too much for current data recording systems to handle.

This means we have to allow the data logging software to decide what to keep and what to throw away during the recording process, before an astronomer gets to look at it.

This is where artificial intelligence (AI) comes in. Over the last years, huge progress has been made with computer learning.

It is becoming possible to have a system learn how to search through huge amounts of data as it comes out of the telescope, filtering out what people have asked it to search for, and also to flag things that are potentially interesting.

Since we have to "teach" the AI systems what to do, there is still a danger of missing something that is really unusual. As the robots do more and we do less, at what point will they deserve to be co-discoverers?

• Venus lies very low in the sunset glow.
• Jupiter and Saturn rise around midnight. Jupiter is the bright one.
• The Moon will reach Last Quarter on the first.

This article is written by or on behalf of an outsourced columnist and does not necessarily reflect the views of Castanet.