D-Wave quantum computer solves protein folding problem | Hacker News Reader

D-Wave quantum computer solves protein folding problem

blogs.nature.com

46 pointsrabidsnail16 comments

> 10,000 measurements using an 81 qubit version of the experiment gave the correct answer just 13 times

> It’s also worth pointing that conventional computers could already solve these particular protein folding problems

So if one cannot already solve the problem is one capable of checking that the answer is correct with such a low probability of correctness?

Protein structures can also be resolved via other methods such as x-ray crystallography of crystallized proteins. Unfortunately, that process can be rather time intensive and isn't viable for all proteins.

Depends on the problem. Factoring into primes seems to be the canonical example.

There are many cases where you can verify an answer much more easily than finding it. Not saying that necessarily applies here, but for the more general problem space it isn't the limitation it might appear to be at first.

We would need to know the distribution of incorrect answers. If it didn't give any other answer more than twice then you could trust the 13-time answer to be the correct output of the computation.

So they can fold a 20 residue protein on a lattice using the coarse grain HP model, competing squarely with 1970s tech.

I read this article and imagined computing in the first half of the 20th century. Undoubtedly when some programmer first managed to get his room sized computer to add, subtract and multiply someone was present to scoff about how he could already do such things on his slide rule.

It's not about where the technology currently is, but where it is going.

All true. With D-Wave, it's always been hard to separate the hype from the substance.

They certainly have some exceptional people (e.g., the "director of business development" mentioned in the post was a research assistant for Hawking (yes, that Hawking), and wrote one of the very first books on quantum computation). And, they're doing good work. It's just hard to tell how long the road is.

Hard not to scoff at with that sensational title.

The phrase for this is "proof of concept". Potentially important research isn't always dramatic, or abrupt.

When I first heard of D-Wave I thought for sure they were some kind of scam outfit. I am impressed..

Same here. That's pretty cool to see that this is actually working.

What's funny is that D-Wave is actually taking the smart approach. They're not trying to build a general purpose quantum computer, they're trying to build an architecture to solve a particular class of problems using quantum computing, which is far more achievable and more likely to be immediately useful. So even at the start they were far less snake-oily than other outfits.

It seems that 13 out of 10000 tests is on the level of noise statistically, isn't it? Admittedly, I am a layman on this subject. But if I had my corgi puppy nose my keyboard 10,000 times a word or 13 would pop out. That doesn't mean she knows how to write.

That very much depends on the size of the solution space. If they get one bit of 'answer' from each qubit used in the computation that's 2^81 possible solutions, in which case you'd expect random chance to require somewhere around 1,000,000,000,000,000,000,000,000 trial runs before they happened to get one right.

Now, I highly doubt the solution space is anywhere near that large. But without knowing more about it we cannot infer anything from the 13/10000 statistic presented.

Am I missing something? There was nothing about the speed in which the correct answer was arrived at versus conventional methods. Isn't q-computing supposed to be significantly faster? Yet, no solution time(s) mentioned.

Reply on news.ycombinator.com