Hacker News new | comments | ask | show | jobs | submit login
Launching the Quantum Artificial Intelligence Lab (2013) (googleresearch.blogspot.com)
50 points by ot on Jan 19, 2014 | hide | past | web | favorite | 28 comments



at first I read this and immediately thought that google had jumped the shark, but then when I realized that they are proposing to use quantum computing to make ANNs tunnel through local minima in converging on solutions, I think that's a really interesting use case - if quantum adiabatic annealing can be easily programmed to do this, (whether or not it has any sort of biomimetic relevance).


Posted 16th May 2013


Anyone here able to recommend a primer on quantum computing? I have a high-level understanding of the physics involved, but I get lost as soon as people start discussing the algorithms.


Drop me an email and I can send you 20+ short PDFs worth of lecture notes from a quantum information course I took at Imperial College, if you wish.

They're probably as good a 'primer' as any, since they basically go from undergraduate quantum mechanics to quantum computing (& other topics). Should certainly make some of the basic quantum algorithms clear (e.g. Grover's).

Failing that, I think Nielsen & Chuang's textbook is really good:

http://www.amazon.com/Quantum-Computation-Information-Annive...

It's thorough, naturally, but it does zoom out from time to time.


Drop me an email and I can send you 20+ short PDFs worth of lecture notes from a quantum information course I took at Imperial College, if you wish.

Any chance you'd be willing to just put that stuff "out there" somewhere, link to them all in a blog post, and submit that here? I'm guessing there would be more than a few people who would find that interesting...


I'd love to. I'm just not quite sure where lecture notes stand, legally speaking.

If I send the files to just a few people I theoretically know, I guess it's like lending a book to a friend. But putting them all up for public download feels like quite a different matter.

Thoughts?


I'm not really sure. I've always tended to assume that if you wrote notes, paraphrasing content that you sat through a lecture on, then you own the copyright on those notes. But I do think I recall at least one case where a professor or a university tried to stop a student from distributing notes, but I don't remember the details...

Here's a couple of articles on the topic:

http://www.wired.com/threatlevel/2008/04/prof-sues-note/

http://blogs.kqed.org/mindshift/2012/02/do-students-have-cop...

http://arstechnica.com/tech-policy/2008/04/aclass-of-copyrig...

Unfortunately, IANAL.


This computer won't let you do "quantum computing" of the kind that's usually discussed - eg, you can't implement Shor's algorithm for factoring. It does claim to be able to solve optimization problems via "true" quantum annealing / tempering, so you'd want to look up what kinds of problems are particularly well suited for those optimization approaches.

http://en.wikipedia.org/wiki/Adiabatic_quantum_computation

http://en.wikipedia.org/wiki/Parallel_tempering

http://en.wikipedia.org/wiki/Quantum_annealing


Check this out: http://www.cs.berkeley.edu/~vazirani/algorithms/chap10.pdf. Vazirani and Dasgupta's Algorithms textbook is amazingly elegant and concise; since Vazirani is an expert in quantum computing, he included a chapter on the fundamentals of quantum in his book!


I highly recommend the quantum computing MOOC from UC Berkeley. I took it when it was on Coursera but now it's on edx.

https://www.edx.org/course/uc-berkeleyx/uc-berkeleyx-cs-191x...


+1, excellent course. I took it via Edx and really enjoyed it.


A good book on the topic for non physicist is schrodinger's killer app published last year. I found entertaining and informative. D-wave computers are adiabetic not gate based and are generally thought not to have significant numbers of entangled particles, this does not allow for exponential speed up that is required for Shor's but should allow for quadratic speed up.


The consensus seems to ever tipping against the idea that D-wave is a true quantum computer. I wonder if Google and others regret spending $10mil on the machine.


It's irrelevant if it's "true quantum"; as long as it can solve optimization problems quicker or cheaper than alternatives, it could be powered by pigeons for all they care.


But if it's not a QC, it won't solve them any faster. Our classical computers are pretty efficient these days.


But it can't.


really? if anything since I first heard about D-wave everything i have heard is in support of their claims. especially the last year or two


Recently in HN: """Comparing the performance of the device on random spin glass instances with limited precision to simulated classical and quantum annealers, we find no evidence of quantum speedup""" http://www.scottaaronson.com/blog/?p=1643


They've had a pretty good PR engine :-)


A little bit off-topic: Aren't quantum computers similar to the 'proposed' PCMOS analog computers based on memristors and spin torque oscillators?

The programming model looks the same, and the machine code is 'complex', so it needs to be abstracted just like in proposed probabilistic computers like DARPA UPSIDE and Synapse?


I think they ought to host a seminar with Hameroff and Penrose like really soon.


I haven't looked into this for a while, wasn't this quantum consciousness stuff widely considered a crackpot phase from an otherwise brilliant scientist (Penrose)?



I recollect this being posted earlier.


so, what are the things that a d-wave system cannot do that a 'pure' quantum computer can?


Solving certain problems faster than a classical computer.


So how worried should I be about this? Are quantum computers really powerful enough to give us strong AI anytime soon?


The problem with quantum computers is that you never know whether they are one or not.




Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: