Computers in Astronomy, really?
Computers have infiltrated almost everything. Something that frightens him is that they have computers in the cars and they all have to work together. The fact that he can get to the grocery store is a minor miracle.
Computer Uses
Controlling Equipment – In this context you might think of telescopes.
Automating Repetative Tasks – This is how computers started.
Organizing and retrieving data – This is our bread and butter.
Building/Exploring Scientific Models
Controlling Equipment
The first computerized telescope he saw, which wasn’t the first one, but it was revolutionary to him, was the, LX 200. It tracks the stars as the earth rotates. It’s very cool, but isn’t good enough to do simple photography.
Vehicle control – We’ve been doing fly by wire technology for decades. He flies small planes for a hobby and when he pushes a pedal, it moves a cable which the rudder. A military plane, which requires constant attention, uses a computer to translate pushing the pedal into moving the rudder. Not direct control.
Old Technology
Feedback technology was originally developed for manufacturing. The roots of “cybernetics” a la Norbert Wiener. These can be sophisticated and adaptive so that the computer drive a robot hand to pick up a can and crush it,then turn around and pick up wine glass.
More Old Technologies
Telerobotics (Waldos)
Pilots in the MidWest fly their planes remotely, in Afghanistan. That’s really phenomenal. They are flying by pure instruments, without haptic feedback.
Unmanned Vehicles
Autonomous vehicles are coming and having successful field trials.
Latest Technologies
Mars Rovers.
Run on RAD6000 boards, which are essentially radiation hardened pcs.
AI and The Mars Rovers
The rovers cannot operate during the martian night, because the computers can’t work at those temperatures. They shut down.
During hte night, human “drivers” decide the rover’s next missions
Build a sequence of commands, run simulations to make certain that they will, send them to the rover.
To Ruben this is frightening, because they write custom programs every day. One bug and no telling what happens. But this is the best compromise, since they can’t control it in real time.
When they sent the robots the software wasn’t ready, so they uploaded it after launch. It makes him shudder, “they are living on the edge.”
The next generation of Rovers will make the simple choices locally, like “hey, this is deeper than I expected, I’ll deal with it” and the people on Earth will decide the high level goals like, “Go get that rock.”
He looked at the resumes of the drivers. They all have advanced degrees. One had degrees in computers, paleoanthropology, electronics and robotics.
Automating Tasks
We depend on automatons for tasks that are:
- too important to be left to (Careless) humans (like temperature control)
- Too boring for (most) humans (sorting records)
- too time-intensive (crunching numbers)
Searching for Comets
Old-school comet hunters scan the sky comparing what they see with a mental map. They’ve memorized it,because if you have to take their eye off the eyepiece to compare it to a sky chart, then they lose valuable time. It takes on average 400 hours to spot a new comet.
This is pretty hard to do well, but many amateur astronomers enjoy it. It’s a competitive “sport.”
Automated Searches.
- This search is a perfect candidate for automation.
- The computer stores a “mental” map of the sky.
- It “looks” at the region each day
- It can determine if anything changed (moved) or if a new “fuzzy” object appears.
He confesses that it takes a lot of the romance away for him. But you can do the same trick for other things, such as supernovas.
Traditional approach is to “blink” two photographic plates of the same location to see any visual differences. Computers can do this continously. Now he says that it is easy, but to compare to images is actually fairly complicated.
Humaning Tasks
Some tasks (even boring ones) are still better left to humans.
The Galaxy Zoo, for example, uses human power (distributed via internet) to classify galaxies. That’s what grad students are, but you do run out of them. That’s what the internet is for. ((Laura says that it’s like a video game PLUS you’re doing real science.))
Data Processing
Data Processing is what made computers popular and affordable. Dull, but important. ((That’s what he said, not me.))
Astronomical Data
A lot of stuff is still in Fortran. Astronomical survey data, covers a wide range of sky in different wavelengths. Sloan Digital Sky.
So how do we make the data available? Sloan data is available through Sky Server. It includes tutotials and educational materials. It also hosts images from the survey. To the Pros they give the data in digital form.
- Spectral
- photometric data
- Spectroscopic data
The search facilities are vital. That’s what’s hard. You have the data but if you can’t find it, then it is useless. Users can search the data using the frontend tools or by providing the raw SQL commands. SQL is OK. It’s from the 70s, like go-go boots, but we want better ways of searching the data.
The problem is that it puts a lot of responsibility on the astronomer to write fast queries, but writing code isn’t their specialty.
Astronomy with Google
The next generation of survey telescopes will push the limits of data storage.
8.4 meter telescope
30 Terabytes of memory per night. To rock your brain, that’s almost 1/2 of the Library of Congress EVERY NIGHT. 1/20th of YouTube
Google is partnering with LSST to provide IT support. (Microsoft works with Sloane)
This is not the first foray into scientific data. They have already volunteered to host terabytes of scientific data, and to make them available to other scientists and the public. Data is sent faster by sneakernet. ((They ship a three terabyte array to astronomer, who uploads their data and sent it back.))
The original idea behind Google
“Better” web pages have lots of links into them.
Note: the hard part in building a search engine isn’t so much finding matches but prioritizing the matches.
Google File System
Uses commodity hardware. PCs running Linux. Lots of Internet bandwidth, arranged in groups (with superfast network connectivity within each group)
Map Reduce
Map operations that do something to the data and then reduce operations that reduce the data and they have lots of these running in parallel. “The internet is great. We have five or six copies laying around the office.”
Google is Academe
Stanford has drunk the coolaid and now teaches MapReduce to their students. Cloud Computer or Utility Computing.
They’re going to far.
Chris Anderson suggested that Google is the end of theory. With petabyes of data running loose, it becomes harder to make any overall sense of the dataset. But there’s no need! GoogleTech lets us sift through this data and find useful relationships.
Google, IBM, and the NSF are building the Cluster Exploratory.
- 1,600 processors.
- TB of memory
- PBs of disk
- Software: Tivoli, GFS, MapReduce
Projected use includes modeling the human brain.
Personally, he thinks this is no replacement for theory and/or models
There is room for model-less science
But the end result should be an understanding of the world.
Ideal sequence: Brahe, Kepler, Newton.
Brahe was a cataloguer. That is data collection.
Kepler looks at the data and is able to infer things from things. he didn’t know why, but he saw the patterns. Kepler science is the Google science. Searches.
Newton is the next step. He looks at these things and says, “I can explain these things.”
Models strike back.
In the 19th and 20th century, models were complicated equations. Many equations are too hard to solve analytically.
The solution is to “integrate” these models by simulating them in a computer. That is what super-computers are doing 90% of the time, solving really hard differential equations.
Computers can give us new insights and new techniques to advance science.
