Hacker News Reader

Comment by cr0sh | Hacker News Reader

I hope they're doing more than just arguing over whether it works or not.

Assuming there is enough information in the paper to replicate the experiment, then the proper course of action would be re-run the experiment and see whether the counter-hypothesis that this won't or can't work is true or not.

Of course, other groups should also do the same. I just hope this isn't another case of "we can't afford to do a replication of the experiment because that doesn't get us grant money" - which I have read about elsewhere is a big deal right now in scientific research (that is, new research is not being replicated - thus science isn't really being done properly).

This research, if it does turn out to be valid, could or would be world changing.

View on HN

Not that you're completely wrong, but it's worth recognizing that experiments aren't exactly trivial to replicate. There's a (sometimes massive) cost in time and money, and expertise can get unbelievably niche.

It's really not as simple as "just replicate it".

> There's a (sometimes massive) cost in time and money, and expertise can get unbelievably niche.

This being a potential trillion dollar business I wouldn't worry about expediency in pursuit of it's veracity

The promise of future trillions doesn't help when you need, say, 300k right now, not to mention that you can't pull requisite expertise out of thin air.

Again, replication is crucial. It's a Good Thing. It's just not a trivial thing that we can expect to "just happen".

300k is nothing given the potential profit of this idea working. Tesla probably spends more than 300k on R&D every week.

Well unless you are going to fund it personally, there are hundreds of universities with the capability and the corporate relationships and the fiscal motivation to do it. Furthermore there are a lot of R&D labs in this sector with the capability in private hands.

The point was this: the paper was _just_ published and people seem to forget that securing funding takes some non-zero amount of time.

I've seen discussions about replication a lot. Isn't it more often than not the case that replication is made expensive because the paper describing it simply leaves out too many details, making replication more of a guessing exercise than actual, well, replication?

"...the paper describing it simply leaves out too many details,"

... and that would be because the patent hasn't been issued yet.

Well, yes and no. Yes, to a certain extent papers don't have enough details for replication, but this is mainly because you just can't fit enough details into the scope of a journal paper. In a PhD thesis, you should be able to.

And no, it's not that replication becomes a guessing exercise, but it's the fact that the time spent setting up, debugging and then running a replication of even a perfectly described benchtop-style experiment costs much much more than the actual equipment involved. And the ROI is poor, since replication studies are hard to publish.

Should, sure.

But there can be critical details needed to replicate which the experimenter didn't realize were important and didn't think to record. And possibly didn't notice.

One of the reasons why replication matters is that it is a way of discovering these factors which matter but weren't obvious.

There's also skill. Its not as simple as running the same shell script twice, this time on your lab's cluster.

So ask the original researchers for more details? What is the problem here?

The problem is that they probably don't want people to be able to easily replicate it, if it really works. It's the same reason compsci papers rarely include source code.

If they don't want people to easily replicate then why do they publish at all? Doesn't it defeat the scientific process? If it's about protection of the invention then why not patent it?

I mean, there is one of these things on a lab bench somewhere.

wouldn't hooking it up to a charge controller, weighing it, then watching how long a lightbulb stays on suffice?

Even if it's a complete black-box, and the thing never gets cracked open, if i understand correctly, the claimed energy density has not been achieved?

If you're replicating a study with your rig, you aren't doing something else with it.

Moreover, replication by the same team isn't what's usually understood by "replication". A large part of the point is that others confirm a scientist's findings.

I agree and really want more real scientific studies but, and perhaps I'm naive here, if they have made a battery which behaves as they describe, you shouldn't need to use any sophisticated methodology to test it. If it looks like a duck, etc.

Supposing it was not a hoax, it becomes a curiosity how it works, and hopefully lead to awesome discoveries, but being able to produce a thing which is great and we don't understand fully is fine by me.

Reply on news.ycombinator.com