So much for science. But at least we can get drunk!

A reader writes:

From: Ernest
To: dan@dansdata.com
Date: Fri, 28 Dec 2012
Subject: Wine Clip

Your review is entertaining but is not very helpful. I have used a Wine Clip for about 3 years, and frequently do a blind test on myself. Naturally, having some one else do the pouring.

After all, I don't really care whether some one else thinks the wine is better using the Clip. I am the only relevant person involved. And it works every time! Red wine is better when the Wine Clip is used. That is what is important to me; after all, I am the consumer.

There is a factor in the use of panels. The participants should never know it is a test. (I found this out when I was doing panel responses. Whether it be taste, smell or any other perception, translation by the brain of the perceptions received is influenced by the environment.

But you are right about one thing. Great magnets!

The Wine Clip

What would you do, Ernest, if someone sent you an e-mail that said, "Your commentary on the Psychotronic Money Magnet is entertaining but is not very helpful. I have used a Money Magnet for about 3 years, and always make more money when I have it hanging around my neck than when I leave it in the bedside drawer"?

Would you, in response to this, drop everything and dash out to buy a Money Magnet from one of the... differently-cognitive... people who sell them?

Would you turn your whole comprehension of the world upside down, because apparently it seems that the free will of other humans and the very workings of abstract probability can be distorted by a talismanic device that works by, uh, quantums, and stuff?

Or would you, rather, presume that the fellow e-mailing you might have not quite the right end of the stick?

I do not doubt that you believe the Wine Clip works. I am intrigued by your claim to have tested it in a controlled, though not double-blind, way. I do not consider your claim plausible, though, not least because if it's correct, then the whole of electrochemistry, indeed most of modern physics, is not. Countless carefully-assayed chemical mixtures are exposed to magnetic fields from the modest to the monstrous every day, with the assumption that those fields will not modify any molecules and mess up the experiment - and the fields never do.

Unless you stick some magnets on a wine bottle, apparently. Then, suddenly, physics goes out the window and "tannins" start getting broken up by magnetism.

Since I am surrounded every day by evidence that magnetic fields do not pull molecules apart, and a good thing too or writing this piece would almost certainly have killed me, I am afraid I can only conclude that there is probably something wrong with your testing regimen. Your collaborator is accidentally signalling you - without your conscious knowledge - or the Clipped pour is consistently the first or the second, or any of the hundreds of other possible variables for which a good test must control, which is why good tests are so difficult to do.

Note that James Randi told me, personally, that he has specifically requested that makers of magnetic wine-treatment devices demonstrate the truth of their claims in return for worldwide fame and a million dollars.

Not a peep.

(You are of course welcome to join the many other believers in paranormal events who say that the Randi Challenge is clearly some kind of scam. I would venture the opinion that a scam-challenge looks more like this.)

But wait a minute - why am I bothering to say all this to you, when you conclude by saying that tests that people know are tests aren't useful anyway?!

I'm currently writing a piece in response to yet another example of audiophile weirdness, and this "the participants should never know it is a test" thing comes up there, too.

I even managed to find someone claiming that the fact that blinded tests are objective makes them bad. Because, see, if a proper test shows you that a $900 audiophile widget does nothing, and you therefore save some money and don't buy that widget, you then won't be as happy listening to music, because even though you now know it was a placebo, you still need that placebo in order to fully enjoy the music. Or something.

But... didn't you say you frequently do blind tests on yourself, Ernest?

If objective testing doesn't work if the testees know they're being tested, I suppose you don't know when these blind tests are happening, right?

So is it something like, your wife sometimes doesn't use the Clip when you think she is using it, or something? And then asks you what you think of the wine, which for some reason doesn't alert you to the fact that another "test" is in progress? And then you turn out to be the first person in human history completely immune to cues from a non-blinded researcher with whom you have a personal relationship?

I'm really trying to not insult your intelligence here, Ernest, but you're not making it easy for me.

Constructive criticism

A reader writes:

From: Al
To: "dan@dansdata.com"
Date: Sun, 23 Dec 2012 14:25:11 -0700
Subject: A little knowledge

I read your Dan's Data on the subject of power chips and must say you have lots to say about a subject you know nothing about. You state that engine timing before top dead centre will cause problems and can destroy the engine. Fact is every engine operates with timing set to vary from 15 to 5 degrees before top dead centre depending on load and rpm. I don't know where you pick up your information but do some serious research before you stick it on the web. I do agree chips are a bunch of hokey.

You're quite right, Al; I've fixed the article. Thanks!

I am gratified to note that now that I have fixed the one mistake you could identify on that page, I have presumably graduated from knowing "nothing" about this subject, according to your intriguing system of evaluation, to knowing everything about it. Will you send me my Ph.D in Automotive Engineering right away, or do you require a small processing fee?

(It's also interesting that that guy from the chip company who offered me handsome remuneration in return for writing a "white paper" about their products didn't notice the mistake either.)

Today's BattleMech advice column

Someone I actually know in real life writes:

I've been following your blog, and decided to install MechWarrior Online to have a go at it (having played the board game version back in '88, the PC CGA version in 91, the new VGA version in 95.... these mechs are all the new ones... bring back the Marauder I say!). Anyhoo... I've played 50 or 60 matches now, and my kill/death ratio is about 0.15 and I have no idea what to be spending my credits on, I'm trying to get a Hunchback build that is reasonable... any tips?

J

Dapper 'Mech

Yeah, that's right: I've got a pinstriped Hunchback, baby. Most important part of the build.

But given the current structure of the game, I think (relative) beginners are best off in a 'Mech that can fit electronic countermeasures, which no Hunchback can. An ECM module largely prevents you from being the focus of enemy attention, and lets you sneak around capturing things in Conquest mode, and you can help teammates just by standing around near them.

Only four 'Mechs so far can accept ECM, though; this Commando, this Raven, this Cicada and this Atlas. Of those, I'd recommend the Raven, because it's pretty fast and flexible, and a 35-tonner, the heaviest possible "light" 'Mech.

This matters, because the current game matchmaker matches any 'Mech on one team with any other 'Mech in the same weight class. So if you launch in a 25-ton Commando, you are likely to attract a Raven or Jenner on the other team. Likewise, Cicada pilots attract Centurions and Hunchbacks.

The new "cadet bonus" feature means new players can actually quite easily afford even an Atlas after not terribly many games (everyone else just got a lump-sum payment of almost eight million C-Bills, which was nice). But if you've already blown that money, the alarming purchase price of the Raven 3L (because it comes with ECM already installed, and an XL engine...) will force you to play quite a few games to buy one. (Or spend eight to ten bucks of real-world money on "Mech Credits" to buy one directly.)

The Commando 2D, on the other hand, costs less than 1.8 million C-bills (or only 715 Mech Credits, about $US2.50 worth). If you can afford the 'Mech but not yet ECM to put in it, you can just run it ECM-less and grind up some experience points while you save money. One solid hit from a big 'Mech can rip off a limb or kill you outright, but as long as you keep moving (which means not running into dead ends and rubbing on walls...) it's surprising how seldom that happens.

The 2D was the Commando everyone feared before ECM, because it's got three missile hardpoints so you can put three Streak SRM launchers in it and hit very surprisingly hard. ECM makes Streak-monsters much less dangerous... unless the Streak-monster has ECM too, in which case he just hits J to switch ECM from disrupt to counter mode, blows away the ECM-packing enemy, then returns to disrupt and keeps on fighting. My 2D has three Streaks, one small laser, and a pretty good kill-to-death ratio.

To answer your actual question, though, I think the secret of a non-annoying Hunchback is to put a big enough engine in it that you can do about eighty kilometres an hour. Then a couple of LRM 5s, some medium lasers or twice as many small lasers (the nine-small-laser Hunchback-4P is a sight to behold), and away you go. Put the lasers in chain-fire mode (backspace, by default, while the appropriate weapon group is highlighted in the bottom right of the HUD), for the best chance of hitting. Chained weapons will fire in slow sequence while you hold the fire button down, but if you want to fire them all quickly you can just click rapidly and get one shot per click.

Beyond that, the difference between frustration and misery and numerous kills and happiness is all in the piloting. Especially if you, like J and I, live in Australia and so routinely get 200-to-300-millisecond pings. When shooting at enemies with lasers with a high ping, ignore the glowing armour your game client shows you, and look at your target's status display at the top right of the HUD. If there are bits of it flashing, you're hitting it (well, someone is, at any rate...). If nothing's flashing, you need to lead the target more.

I'm sure that even you, J, will eventually be able to master this, despite the miserable reflexes and poor concentration that no doubt got you into your cushy government job driving a steamroller or a tram or whatever that thing is you actually drive at work.

Oh, and with regard to the abovementioned various versions of this game... It's come quite a way, hasn't it?

UPDATE - J's reply!

Sweet... I've spent most of tonight running around in a Raven 3L :) Still have to work out the best way to use the TAG and NARC... maybe I'll upgrade the engine to get more than 97kph out of it.

NARC launchers are not much use in the game as it stands. The last patch boosted the duration of a NARC beacon from 15 to 20 seconds, at least, but 20 seconds still isn't very long, and ECM neutralises NARC completely.

And the darn launcher weighs three tons, requires ammo, and that ammo only gives you six lousy missiles per ton. (This makes NARC missiles the heaviest ammo in the game; even AC/20 ammo gives you seven shots per ton!)

TAG, on the other hand, is only one ton (a NARC launcher is three tons), needs no ammo and makes no heat. So it's not much of a commitment to install it, and wave it around gaily like an overstimulated raver at almost all times. The last patch also increased TAG range from 450 to 750 metres.

You need to keep TAG-lasering a target to be of any use to your team, though. The TAG effect lasts for three seconds after you cease illuminating the target, so you can keep flicking the beam over targets and at least don't have to hold it on them constantly. But brief drive-by TAG-ing is worse than no TAG-ing at all. You'll just have the missile-boats on your team lining up a shot on your briefly-illuminated target and then, usually, launching precious missiles a millionth of a second after the target lock disappears again.

If you're running among the enemy being an ECM nuisance, you're going to have trouble consistently illuminating them; if you're TAG-ing them from a distance you're announcing your location, which can be helpful if you're devoted to Team Annoying Bastard tactics, but will usually just get you killed.

For newbies, this video shows how to use TAG - it's the barely-visible red beam coming in from the right side of the cockpit view.

With regard to the engine, the biggest motor you can jam into a Raven 2X or 4X is a 245, giving 113.4km/h. The Raven 3L can take engines up to 295, though, giving a speed of 136.5 km/h. That's the fastest any 'Mech can go in the game thus far - apparently speeds above 150 km/h currently cause the game engine to do weird things, and the "Speed Tweak" elite upgrade adds 10% to your base speed, which takes a 136.5-km/h 'Mech neatly up to 150.

Even an XL 295 engine weighs 15 tons, though, making it difficult to cram much in the way of weapons into a XL-295 35-ton 'Mech like the Raven. Personally, I'm quite happy with the stock XL 210. If you want lightning speed with some room for weapons, the abovementioned Commando 2D will give it to you; my ECM Streak Commando has an XL 195 in it, giving a base speed of 126.3 km/h.

You can also use your Raven to explore different play styles and get the hang of big 'Mechs without having to buy one. Install a small engine, giving you the speed of a heavy or assault 'Mech, and some appropriate fraction of that 'Mech's weapon loadout, and then hang around with the heavies and do what they do.

I had a lot of fun in the pre-ECM world with my slow Raven missile boat, which was essentially half of a Catapult.

Just don't install a big missile rack in a location which, from the factory, had a NARC launcher; those locations usually have only one missile tube, and so wee out the missiles one at a time. This is a neat way of sucking all of the AMS ammo out of an enemy, and it's funny to watch, but that's as complimentary as I can be about it.

Free ammo!

I was writing something about ammunition expenses in Mechwarrior Online (each individual Artemis Long Range Missile costs 334 C-Bills; each Gauss round is 2000!), but then Piranha Games had to go and ruin everything with this news.

As of the eighteenth of December (which it already is here ion Australia as I write this, but isn't quite yet in Vancouver, where Piranha are), repair and re-arm costs are going away, completely.

So install XL engines and expensive-ammo-ed weapons to your heart's content! You'll still have to pay for ammo when you first install the weapon, I think, but no longer will you be installing extra ammo bins so that the free 75% reload after a game will give you enough to get along, without paying a fortune for that last 25%.

Newbies will get a lot more money - a "Cadet bonus" - for playing their first 25 games, so they can afford a reasonable 'Mech quickly. And trial 'Mechs will now earn as large a C-Bill and experience point reward as owned 'Mechs. So when you buy your first 'Mech you may already be able to afford an efficiency upgrade or two. You still can't upgrade or modify trial 'Mechs, but now they'll just be normal 'Mechs with I presume the traditional lousy setup, rather than complete second class citizens.

Oh, and non-newbies will be getting about eight million C-Bills, the total Cadet bonus, just as a present.

Rewards for starting a game and just standing around doing nothing are now very small. Rewards for killing enemies are much higher. The money and XP reward for assisting in a kill was until now the same as the reward for kill; now you will actually get quite a lot more of a reward for assisting than for being the person who fires the killing shot!

Also, rewards for capturing the enemy base in Assault games are now zero. So nobody will be shouting at the rest of their team to leave the last AFK baddie standing there and capture instead. Capturing is now just a way to end a game when you can't find, or don't have time to get to, the last enemy.

(You will of course get rewards for capturing things in the upcoming Conquest mode, since that's the whole point of that game mode.)

Further 'Mechitude

Penny Arcade produced a mildly amusing thing for MechWarrior Online. It's OK, but it ain't no Vault 77.

(To which there does not seem to be any good way to link. Click that link, and lie about your age, and then click the link again, and then click from the final page it helpfully shows first back to the first page...)

More importantly, if you click a button on the MWO Penny Arcade page, you can get a free day of "Premium Time".

(If your browser pops up a malware warning when you go to mwomercs.com, by the way, it's probably a false positive not a false positive, but the problem has now been dealt with. The whole concept of the hyperlink seems to be disintegrating today.)

Premium Time gives you a 50% boost to all money and experience points earned over the premium period, but once you "activate" it in the game client, you can't de-activate it again. The timer ticks down to zero over the allotted time, whether you play the game or not.

The most sensible way to use the free 24 hours is therefore, clearly, to kit yourself out with intravenous feeding and a bucket to sit on and play non-stop for a straight day over the weekend. Or, indeed, during time when you're meant to be working, studying, sleeping, collecting your children from school, governing a nation, or what-have-you.

I haven't written about MWO for a couple of weeks. There've been some changes, the most important of which is the introduction of Electronic Countermeasures - ECM.

The 1.5-ton Guardian ECM unit can be installed on exactly four 'Mech variants, so far; one model each of Commando, Raven, Cicada and Atlas. In essence, it creates a 180-metre-radius bubble around the 'Mech it's on, in which no enemy can get a target lock, unless an enemy ECM-equipped 'Mech in the same range has switched their own ECM from the standard "disrupt" mode to "counter".

There's a bit more to it than that; here's a lengthy screed on the subject. (Here's the same thing on the official forums, which as mentioned above currently pop a malware warning for me, in Chrome.)

When ECM was new almost everybody used it, which is what happens whenever they introduce a new feature. This made it pretty much pointless to launch in a 'Mech that relied on long-range or Streak short-range missiles, because you'd almost never get a lock, even if you had your own ECM with which to counter one of the horde of enemy ones.

Now ECM is no longer new and exciting, and it's become normal for pick-up teams to contain only a couple of ECM 'Mechs out of eight players, or even none at all. LRM boats are now quite useful again, though no longer the relaxing play experience they used to be. And the Scourge of the StreakCat...

...has been ended, along with the massive dominance of the Jenner and to a lesser extent the three-Streak Commando in the light category.

It's still wise to run an ECM 'Mech in pick-up games, and it's not difficult to get into those, at the low end. A Commando-2D will only cost you 1.8 million C-Bills, and a Guardian unit is only 400,000 on top of that.

(You can also buy 'Mechs, but not equipment for them, with real money via the "Mech Credits" system. A Commando-2D costs 715 Mech Credits, which is not much. PGI are continuing their merciless war of attrition against their customers' wallets by currently offering a 20% bonus on all of their Mech Credit packages; the smallest $US6.95 one therefore now gives you 1500 MC. That's enough to buy a Commando-2D and a whole other different Commando variant too, to get you started on the efficiency unlock process which requires levelling three variants of a given 'Mech to get to the tasty Elite upgrades. Unlocking Elite also doubles the effectiveness of all of the Basic upgrades, though contrary to popular belief unlocking the final Master upgrade that gives you another module slot does not double the Elite upgrades. Mech Credits are also how you normally buy Premium Time; one day of Premium costs 250 MC, and there are other packages ranging from three to a whopping 360 days. The one-day price is no more than $US1.39 even if you buy the worst-value MC package that exists; 180 days is 13,500 MC, which you could just afford with the current 20%-boosted $US49.95 MC package. 360 days is 24,000 MC, eighty US dollars if you buy the best-value $US99.95 20%-boosted MC package.)

Man, that was a big parenthesis. Here's a palate-cleanser.

Behold, fellow MechWarriors, the Noisy Cricket!

"You might want to have something to do while you wait for your mech to cool down. I watch a stream and enjoy my coffee. Or wash the dishes."

It's kind of like the last sword-fight in Rob Roy: He's only got to hit you once, fancy boy.

The tradeoffs for this thing are terrible, of course. Not only do you overheat and shut down after every shot (you could avoid this by chain-firing (this player has the same four ER-PPCs in two groups, the first of which is chained), but then heat would limit you to the same rate of fire you'd get from only two, or at most three, ER-PPCs. THe massive weight of the guns in this titchy 40-ton 'Mech also means you can only fit a tiny engine, which gives this oversized-scout-'Mech a top speed of fifty. And it doesn't have much armour, or, plainly, nearly enough heat sinks.

But, pew pew!

Of guns with, and guns on, rails

In the comments of my post the other day comparing electrical and firearm energy levels, commenter "hagmanti" was delighted to be informed that the blast of flame coming out of a rail-gun's barrel...

Railgun muzzle blast

...which makes it look more like a normal chemical firearm than most normal chemical firearms do, is vapourised rail and projectile material.

This is a serious problem for both military and... hobbyist... railguns. Damage to the projectile is not that big a deal as long as you're only firing "kinetic kill" lumps of metal, not explosive-filled shells. But a gun that needs to be torn down and have major components replaced every few shots is not a practical weapon.

There are actually analogous problems with a lot of other unreasonably powerful guns. Truly monstrous artillery like railway guns (often, confusingly, also referred to as "rail guns") could fire only a few hundred rounds - even with WWII technology - before the whole huge barrel had to be replaced.

Paris Gun

At least one of those railway guns, the World-War-One Paris Gun, had a series of shells of gradually increasing size to be shot in order, so the bullet always fit the barrel.

The same thing happens to small arms. A frequently-used rifle barrel will eventually be "shot out" and lose projectile velocity and accuracy, as the bullet bounces down the worn tube. It just happens a lot faster if you perversely insist on 120 million joules, or around 1.7 billion joules, of muzzle energy, for the Paris Gun and...

Schwerer Gustav

...Schwerer Gustav, respectively.

For comparison, the 16-inch guns on the Iowa-class battleships were good for a feeble 355 million joules or so, and three hundred or so full-power firings before the barrels wore out.

In their later life those mere 16-inchers became rather more destructive than any of the railway guns, though, on account of how they could toss fifteen to twenty kilotons of instant sunshine at the enemy.

And then there's the multi-chamber gun concept, where the initial propellant charge behind the shell is relatively small, and the barrel has branches containing subsidiary charges that are timed to go off after the shell has passed them. This design lets you have very high muzzle velocity without beating up the barrel, or the shell, with a single immense propellant explosion; multi-chamber guns could be used to launch satellites, as well as to kill people. But nobody's ever really gotten them to work, which is, I think, in most cases just as well.

Building a better Bond

I'm a bit disappointed in the recent semi-rebooted gritty James Bond movies.

They're good films, and they're far better than the burgeoning silliness of the last of the old run of movies (...an invisible car? Really?). But a lot of that silliness was just misapplication of one of the hallmarks of the classic Bond movies: Gadgets.

The gadgets played a large role in making Bond films what they were, but they sort of stayed phase-locked in the Seventies. Bond might have been remote-controlling his BMW [shudder] with a Nokia or something, but the Third Doctor had a frickin' remote-controlled car in 1971. Bond didn't even have a gadget with a cutting laser on it until 1983.

If you're going to have Bond gadgets again beyond the low-key stuff in the rebooted films, you have to make them truly impressive. Not something, like an invisible car, that could have been dreamed up in 1970 as easily as 2002.

This is the secretest of secret agents going on the most important missions ever, after all. He should be kitted out with and backed up by with the very best superblack reverse-engineered-from-crashed-flying-saucers ultra-technology that can be created by the distinguished successors to Bletchley Park (and all the other people whose discoveries went into Tizard's briefcase).

So, say:


Bond has been shepherded into a lift by Q, and they descend. For a rather long time.

On the way down, Q explains that MI6 and, ah, some higher-numbered agencies, rather suspect that certain developments in mechanical augmentation of human strength, for military and industrial purposes, may have fallen into the wrong hands.

And that there is no real reason for these systems to be limited to only a man-sized exoskeleton, or indeed for constructors to tolerate the weakness of a normal human body within it, if one is willing to take certain rather drastic steps to ameliorate this problem.

And that Her Majesty's Secret Services have been working on their own systems to combat this threat, but have faced certain ethical obstacles.

The lift doors open to reveal a warehouse-like space, harshly illuminated by overhead fluorescents, and dotted with computer installations, machine tools, and agglomerations of technology of unclear purpose.

The giant room is dominated, however, by a looming object in its centre. A mad profusion of cables and pipes and screens and scaffolding and catwalks surrounds, and obscures, a metallic shape about the size of a terrace house.

Q turns to Bond, and says, "For this project to succeed, 007, we needed someone with great familiarity with our most advanced systems; otherwise the training process would be impossibly difficult. There were several candidates, but owing to the... the nature of the project, none were acceptable."

"Pardon?"

"Well, not to put too fine a point on it, we needed their brain, and about six inches of spinal cord, which is rather-"

"You needed...?!"

"Which is, is of course, more than we were prepared to ask any servant of the Queen to volunteer. But then-"

"What the hell are you-"

"But then, the previous Q had, well, he had a car accident. And, fortuitously..."

He waves vaguely at the huge shape in the middle of that mass of pipes and cables.

With a subsonic hum, the shape changes.

It stands up.

Many of the cables and pipes drop away, as the giant machine takes a step forward. The concrete floor trembles noticeably as its foot comes down.

The machine stops.

It speaks.

"NOW PAY ATTENTION, DOUBLE-O-SEVEN."

More photons, fewer pixies

A reader writes:

This gravity powered light has been getting a little attention lately. It reminded me of your post on the Gravia light.

So is the GravityLight a sensible design that conforms to the laws of physics, or is it also powered by Pixie Dust?

Lee

Gravity Light

There's nothing inherently wrong with the basic idea of converting the energy of a falling mass into electricity. (That's how hydoelectric power stations work, after all.) It may even be possible to do it quite efficiently, and cheaply, on a small scale now.

In the olden days you would have needed to gear up your dynamo a lot from the pulley your falling weight (or flow of stream water) was turning, but today super-powerful magnets are very cheap and so efficient lower-speed dynamos are easier to make. Especially if you're only trying to light one high-intensity LED to a brightness that'll let someone read a book at close range.

The GravityLight people claim thirty minutes of light per lift of the weight, and the tape connecting the weight to the light looks to be, being generous, about 1.5 metres long. They don't say exactly what the weight is, but they do say you can fill the GravityLight weight bag...

GravityLight kit

...with use "anything weighing about 20lbs"; let's again be generous and say you get ten kilos of stuff in there, which is a perfectly liftable weight for a wide range of humans.

So we've got ten kilograms falling 1.5 metres in standard gravitation; that gives us a maximum of 147 joules to play with.

Split over thirty minutes, which is 1800 seconds, that gives us only 0.082 joules per second. That means a power of 0.082 watts, or 82 milliwatts.

A single standard white LED will have spec-sheet power numbers of about 20 milliamps at 3.6 volts, which is 72 milliwatts. It'll work fine - and more efficiently - at rather lower current, though, which is just as well because this falling-weight system is certain to be a long way short of 100% efficient at turning the weight's gravitational potential energy into light.

Even if it's only 50% efficient, though, you've still got 36 milliwatts to play with, which is plenty to light one LED to a useful, though far from room-filling, brightness. Even if you pare off some of the above generous assumptions about weight and fall distance, you'll still be easily above 20 milliwatts, which is also usefully bright.

The GravityLight people say their invention is intended to replace kerosene lanterns, but it's definitely not going to have the room-filling brightness of a kerosene lamp with the wick turned well up. Given the numerous downsides of kerosene lighting, though, and the fact that a lot of poor people probably don't turn their lamps up to max very often, a GravityLight or two could well replace one.

The ridiculous Gravia concept thing had a heavier weight falling a similar distance, for a total of about 271 joules. But the designer idiotically claimed that the thing would have the light output of a forty-watt incandescent bulb for four hours. This was so impossible that it couldn't come anywhere near being done even in Physics Experiment Land, with a perfectly efficient dynamo and a perfectly efficient lamp.

With real-world hardware, in contrast, the GravityLight can work. I'm not totally convinced that for household lighting you wouldn't be better off with a couple of conductive objects a reasonable distance from each other in the galvanic series, some damp earth as an eletrolyte, and a Joule Thief to boost the output to run an LED. That sort of improvised battery can run for a very long time at the very low power a Joule-Thiefed LED requires, and its poor portability doesn't matter if you're using it as a hoursehold light (and is an advantage if you want to avoid your light being stolen...).

But the physics, at least, checks out for the GravityLight.