Everything you do will help us
I’m Jill Tarter, the Director of the Center for SETI Research. Since we launched at TED last Wednesday, I’ve been reading what you’ve been writing. There have been a bunch of comments on SETILive about not knowing what to do or what to mark or whether you are getting it right. We’ll work on making the tutorial more accessible and more informative as you’ve suggested, and over time we will implement some better marking tools as you’ve requested – but the ‘getting it right’ part is a bit more dicey. That’s because we really don’t know yet exactly what ‘right’ is.
As Lou Nigra (thanks, Lou!) and the tutorials have described, the SETILive data that are coming from the ATA originate in the crowded bands; small portions of the terrestrial microwave window that we have historically skipped over. That’s because our SonATA system gets confused there – it detects LOTS of signals, but it cannot finish clustering them, and classifying them by comparing them to signals that are detected in the other two (or maybe one as is now the case) beams on the sky being observed simultaneously, or finish looking them up in a database of all the signals that have been tagged as RFI in the past week. Rather than conduct our observations with non-uniform sensitivity, or continuously restart software modules that have given up in exhaustion, we have chosen to ignore these crowded bands – at these frequencies we’ve been blind and deaf. Ultimately that might turn out to be the best strategy – after all, why are those bands crowded? They are crowded because they have been allocated to different types of terrestrial communications services. We are the ones making all those signals. Or are we?
IF (of course it’s a huge if) there is a technological civilization near enough to us – its distance in light years is less than half the time over which our technology has been transmitting at a particular frequency band – perhaps that civilization has noticed that the Earth is very ‘radio bright’ at certain frequencies. Perhaps it has transponded back a reply at the same frequencies, knowing that we would have receivers that work there. A bit more speculation suggests that their message may be crafted to be detectable against this background of terrestrial transmissions. With this scenario in mind, we could try to code and implement all sorts of clever, non-linear anomaly detectors that inter-compare the signals received from the multiple beams on the sky – but remember we are trying to do this in near-real-time. The detector has to finish this task significantly before the observations move on to the next frequency band, because the system still needs to match whatever the detector has found against recently detected RFI from other directions on the sky. We don’t know what we are looking for, but we do want to invoke logical constraints that insure that the signal is only coming from one direction on the sky and not many.
Before we throw a whole lot of new computing resources (that we actually don’t happen to have) at this problem, we should take a look at what’s actually going on in the crowded bands as a guide to what might be the most effective strategy – that’s where you come in! We are hoping to use the amazing pattern recognition of your eyes and brains to look for signals (patterns of some sort) that appear in only one beam and not in any of the others. We hope you can help us set up a sort of rogues’ gallery of signal patterns detected over the past week (fortnight, month, 3 days ??) that can be collectively ‘remembered’ to assess whether this particular signal pattern has been seen before from other directions on the sky. That’s why we want you to mark the RFI in multiple beams as well as any pattern that only shows up in one beam. And then if enough of you mark the same single-beam pattern (so we are fairly confident it’s real, not noise), we’ll decide that it’s an interesting candidate signal and follow up on it immediately. That means that instead of moving on to the next frequency in the observing sequence, we will reobserve in the same directions, at the same frequency. SonATA is still blind, so you will have to tell us whether the pattern persists – is it still there? Is it still only in one beam? If so, the next observation will observe at the same frequency, but looking at different directions. Is the pattern still there? Well, that’s too bad, it means it really was some form of interference and isn’t associated with the target we were pointing at on the sky. BE PREPARED – WE THINK THIS WILL HAPPEN A LOT. Just like your eyes have peripheral vision, a radio telescope has ‘sidelobes’ into which signals can scatter and be confused with signals entering from the direction the telescope is pointing. The sidelobes are complicated in the way they cover the sky; it may appear that a signal is coming from only one beam out of three, but moving ‘off source’ can reposition the sidelobes so that the interference is once again detectable.
But what if the signal/pattern persists when we reobserve ‘on source’, and disappears when we go ‘off source’? That’s getting interesting! We’ll start up a cycle of ‘ons’ and ‘offs’ that will stop when the signal fails to be detected, or not be detected, at the right time, or when we’ve completed five cycles. If the system successfully completes five cycles, then the team at the Center for SETI Research will be alerted and we’ll be right there with you using our eyes and brains to figure out what to do next. Since we’ve begun SETI observing on the ATA this has not happened in the less crowded bands that SonATA has been exploring automatically. Now that we are trying to probe the crowded bands, we’ll have to see how it goes.
By now I hope you are convinced that your efforts can only help us. There’s a slight chance that you just might discover a signal from another technology buried underneath all the terrestrial interference and we will all celebrate. But at the very least you’ll help us better understand what it is that humans are doing as they manage to look at complex patterns and isolate sub-patterns that are unique to one of multiple samples. There may well be neurologists or psychophysicists out there who already know that answer, but my team doesn’t. If we can learn from you, we can be better equipped to train future automated detectors. And if it turns out that this is not a task at which humans are particularly adept, well we haven’t lost anything. After all, our previous strategy was to ignore the crowded bands. There is only an up side to your participation.
Thanks for being willing to help out, and good luck!