NASA and Artificial Intelligence
Why Google’s new quantum computer could launch an artificial intelligence arms race, Washington Post
“Ever since the 1980s, researchers have been working on the development of a quantum computer that would be exponentially more powerful than any of the digital computers that exist today. And now Google, in collaboration with NASA, says it has a quantum computer the D-Wave 2X that actually works. Google claims the D-Wave 2X is 100 million times faster than any of today’s machines. As a result, this quantum computer could theoretically complete calculations within seconds to a problem that might take a digital computer 10,000 years to calculate. That’s particularly important, given the difficult tasks that today’s computers are called upon to complete and the staggering amount of data they are called upon to process. On the surface, the D-Wave 2X represents not just a quantum leap for computing, but also for the field of artificial intelligence. In fact, Google refers to its work being carried out at NASA’s Ames Research Center as “quantum artificial intelligence.” That’s because machine learning problems that today are too hard or too complex for computers could be solved almost instantaneously in the future.”
5 things you should know about the plan to open source artificial intelligence, Washington Post
“Arguably, the open source movement the idea that a group of technologists freely contributing their own work and commenting on the work of others, can create a final product that is comparable with anything that a commercial enterprise might create has been one of the great innovation catalysts of the technology industry. It’s no wonder, then, that a group of Silicon Valley luminaries including Elon Musk, Peter Thiel and Reid Hoffman have lined up to contribute $1 billion to a new open-source AI project known as OpenAI that is led by Ilya Sutskever, one of the world’s top experts in machine learning. If you can open-source software and hardware, then why not open-source artificial intelligence, right?”
We need better on-board diagnostic tools than smarter computers on probes, I think. NASA wants to control and examine stuff as a probe goes along anyways, so having it be “smarter” doesn’t really help that much.
I’m not sure what you mean by better on-board diagnostic tools. The bottleneck isn’t the hardware for diagnostics, it’s the telemetry to send it down. Typically housekeeping comes down at very low time resolution, unless something goes wrong and the spacecraft goes into (or is commanded into) a non-science mode. If you’re talking about interpreting those data, isn’t that a smarter computer?
But I see real advantages with smarter on-board logic. My main objection to autonomous rovers is that they would just be autonomously going from one place to another. A tele-operated one (or a geologist in the field) can notice something interesting along the way, decide to make a brief stop and some measurements, then continue on.
A very common problem in my field (space physics, with an emphasis on other planets) is dealing with very high time resolution data. For example electric fields measured at microsecond time resolution (48 Mbps before compression). Typically most of those data are simply nothing special, so there is no need for an unsupportable data rate. It’s only the occasional one-second interval when something interesting is happening. So the solution is often to write the data to a buffer, have the flight software automatically decide which brief intervals are “interesting”, send them down and delete the rest. The flight software currently isn’t all that smart, and can frequently get fooled into capturing the wrong events (not that doesn’t occasionally result in new discoveries, but it can also be a pain.)
I agree with Einstein and Paul. Quantum computing, in theory, uses superposition of quantum particles to make very large numbers of simple calculations more rapidly than is possible with conventional electronics. It has potential applications where supercomputers are used today, in cryptography, computational fluid dynamics and weather and climate modelling. But quantum computing is unlikely to find any application in AI, which requires complex software and algorithms but only modest computational speed.
NASA needs AI. It is the only federal agency that operates robotic systems so far away from Earth that direct teleoperation by humans is not feasible, and AI has major applications in aviation as well. tt would make sense for NASA to work with aircraft and spacecraft manufacturers on development of AI software and algorithms for aircraft, spacecraft and rover applications, and with electronics manufacturers on radiation hardened spacecraft computers that would at least approach the capacity of a conventional desktop.
I agree however, there are other agencies involved in remote systems other than NASA (or DoD). More interestingly, my work on managing other forms of automation and operations as complex systems of systems has an analog with remote sensing platforms for which my former colleagues at GSFC gather space science data. Will I build a platform that a C-suite operations guy/gal can understand where AI will generate sales revenue, higher margins, more efficient value chains, and improved customer relationships? Perhaps these new quantum tools will “lower the cost” of implementation of these systems. Just keep the big consulting companies out of this, please!
Uh guys (Brett/Daniel), the sort of hardware necessary to run a good AI (even at current state-of-the-art) is not only beyond anything you can power in a rover, it is not even vaguely rad-hardened and therefore would last a few hours before being irrevocably ruined.
Computers in spacecraft are generations behind the curve.
Not sure AI is beyond rover power capacity as it does not require a supercomputer and spacecraft computers could be considerably more capable with just some improvements in radiation-hardened electronics.
D-wave is not really a quantum computer as the hype claims and it has nothing to do with AI; AI has nothing to do with intelligence btw– it is just a bunch of ad hoc optimization algorithms constantly needs human fine tuning. As to open source, google has already released its AI (deep learning) package, what did Musk do? Yet Mr Musk is here to steal the thunder, how unremarkable and even boring.