Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!


Forgot your password?

Comment Re:No Alternatives??? (Score 1) 170

It's an engineering problem. You surely could get some combination of solar and battery to work on the Martian surface, but it would impose design and operational constraints -- constraints which could be mitigated with money.

Presumably they crunched the numbers and developing an entirely novel compact reactor looks like it could be a win. However lets imagine this "Kilopower" project is a total failure; that doesn't mean that a Mars habitation mission couldn't proceed, it'd just cost more to get a certain amount done.

Comment Re:Uhm... No? (Score 3, Insightful) 94

It's not as interesting as what I though it was going to be about. I'd like to see a series of a dozen or two small, high-temperature-capable reaction vessels (some glass, some platinum-coated steel), each with its own temperature and pressure regulation hardware, and self-reconfiguring plumbing fixtures attached to it (gas/liquid multiplexers). Some vessels would come with common catalyst packs in them (platinum, vanadium oxide, iron, etc), some capable of maintaining a temperature gradient for distillation, some for gas-liquid exchange, some with stirring hardware or an auger to remove precipitates, one with electrodes for electrolysis, etc. A couple heat exchangers also would be nice (potentially the same hardware as the MUXes), as well as a the obligate pump(s) and compressor(s). And of course you need hoppers for solid feedstocks, feed lines for liquids and gases, etc. A nice touch would be if one or more XYZ-axis arms could move between different feedstocks and/or containers for finished products.

Something like that, where the vessels remain constant but the lines between them reconfigure based on software inputs, would be amazing. Doesn't need to be large - even a desk-sized unit would be very useful. And such a thing would be invaluable for space applications, too; it's one thing to set up offworld production of certain largescale feedstocks, but a whole different thing to try to set up production of every chemical we use as a society, and in particular those needed to keep your industrial processes going. Small-scale batch synthesis is an option, but that requires human labour, and humans leave a massive trail of required consumables in their wake. Automated lab synthesis, however...

But as for this? I don't see the point of the 3d printer. They're just printing a bunch of simply interconnected vessels and then manually doing a series of reactions in them.

Comment Re:the thing about teslas autopilot... (Score 4, Interesting) 155

It's not just about appearances. It's about cost, drag, and power consumption. Lidar is a pain on all three of those (in addition to looks). You simply can't sell cars with big $10k domes bulging out of the top upping your drag coefficient by 10-20% and consuming a couple kilowatts of power. That would be a disaster to your range, and make your vehicle totally uncompetitive.

"More than one year after launching V2, Autopilot still lacks some of the functionality of the original, and there are many anecdotal reports from owners of unpredictable behavior."

Funny how you don't get anecdotal reports concerning the others, given that most of them don't have owners to make said anecdotal reports. And of most of the competitors' systems, they're comically bad. And they have the gall to actually market the car as currently "self-driving" (unlike Tesla which markets self-driving as an additional package which you can buy but won't be active for years).

Some of my favorite quotes from the test drive comparison:

One never really decides to engage Drive Pilot. You press two buttons on the left side of the dash, one for Distronic Cruise Control, the other for Automatic Steering, then press a button on the left side of the steering wheel, then, — when Drive Pilot decides conditions are suitable — it engages.

Is there an audible sound? None that I heard. Like Autopilot, a green steering wheel icon illuminates on the bottom center of the display, and is duplicated in the Heads-Up Display.

Engagement is made clear by the car’s instant and unsafe wandering in all but perfect conditions, and often in perfect conditions.

Unlike with Autopilot, placing your hands on the wheel and steering doesn’t instantly disengage Drive Pilot. I suppose this is intended as a method of allowing the user to guide Drive Pilot by making course corrections, but instead it resulted in an unwanted and stressful upper arm workout, without which I’d have been killed.

I got the Drive Pilot to “drive” itself for as long as sixty seconds, which is as along as Mercedes-Benz deems it safe. Trust me, you don’t want to take your hands off the wheel that long unless your car’s on fire and you’re reaching for a fire extinguisher, and even then.

Drive Pilot had a nasty habit of disengaging in good conditions before sixty seconds were up, with no obvious warning except the green steering icon going out, and lane drift. After the third time, I actually felt fear.

This is actually a dangerous product. The car will steer itself into oncoming traffic. It oscillates between lane markings like a drunk driver. No setting or speed is sufficient to compensate for the utter failure of this functionality.

Did anyone in Stuttgart drive a Tesla on Autopilot? Even once?

People need to be fired. Think I’m being harsh? Here’s another direct comparison between Drive Pilot and Autopilot, from Norway’s Autofil. Scroll down to the pictures comparing the two cars' lane keeping. Need more convincing? Here's Wired's take. Still don't believe me? Video is coming soon, via Drive on NBC Sports.

The only good thing about Drive Pilot is that your Mercedes will protect you from it. Did I trust it? Only at a crawl. Did I understand it? I don’t understand how Mercedes-Benz could release this to the public. I hated literally everything about it. It drove like a drunk ten year old, fighting for the wheel with a drunk fourteen year old.

They've "improved" it since that article, but their main improvement has been a massive cheat: in practice it spends most of its time locked onto whatever car is in front of you, doing whatever they do - including unwanted lane changes, driving like an idiot, etc. Autopilot certainly incorporates the behavior of the vehicle ahead of you into its datastream, but it doesn't simply mimick their behavior.

Some on the market products offer a better autopilot experience by a different cheat: premapping. They drive a Lidar truck down the roads and map the geometry everywhere on that route. So your car stays nice and perfectly centred in your lane and does the curves just perfectly. Assuming that you're on a road they've mapped, otherwise it doesn't work at all. Assuming that the road hasn't changed and that there's nothing the TACC radar can't see in the road, or it'll happily plow you straight into disaster. Assuming that your GPS doesn't glitch out, or.. same thing.

Then you have the "talk a good game but no actual product on the market" companies like Google's Waymo. They have a great safety record!.... in the precise conditions that they've chosen to test the vehicle. Using hardware that is in no way commercially practical to put on consumer vehicles. Etc.

Tesla has taken on the hardest of problems: a real-time (not pre-mapped), commercially viable (not super-expensive, ugly, high drag, power-sucking), massively deployed in the real-world Autopilot system. And yes, they absolutely have had setbacks - the harsh divorve with Mobileye made them switch over to AP 2.0 before it was ready, and they've spent the last year playing catch-up with themselves, only recently starting to exceed where they were in AP 1.0. And yes, the gap with at least some of their competitors is closing. But as for the claims in this report, they're just laughable.

Honestly, I don't think the future is traditional LIDAR. Rather, I think it's time-of-flight cameras. They should be able to be mass produced for a similar cost to visual cameras, with a similar profile, and provide both ranging data as well as the unavoidably-necessary visual data for optical processing. For those currently with visual/radar systems for ranging, time-of-flight ranging data should be a direct drop-in for photogrammetric ranging, which is more prone to stitching errors (photogrammetry avoids stitching errors with our eyes only because our brains reason out what "makes sense"; what "makes sense" is an AI-hard problem)

Comment Re:Duh, missing other data dimensions (Score 2) 161

It tried to be fair and actually failed, because it uses a methodology that clearly wasn't designed by a statistician.

The program uses over a hundred factors in its classification scheme, but statisticians and data scientists make a point of pruning factors because long experience has shown that introducing many irrelevant factors actually reduces predictive accuracy. And just because race is not an explicit factor doesn't mean that the algorithm is race blind either. It's entirely feasible to given the huge number of factors involve to recover the subject's race with a better-than-chance reliabilty, whether explicitly or implicitly; intentionally or even by accident.

Now the program's score is equally correlated with reoffending rates whether the subject happens to be white or black, which sounds impressive and color-blind -- to a layman. To a mathematician not so much. It's actually quite easy to produce this result by tweaking your model, implicitly recovering race in the manner suggested above and forcing it to produce a result that looks right -- in aggregate.

But what a statistician wants to know is about conditional probabilities, and it turns out that when applied to retrospective data the program is twice as likely to commit a type 1 error (falsely predicting reoffending) for black subjects as white. If this makes the whole process of achieving fairness sound hard, that's because it is. Color-blindness in aggregate isn't the same as color blindness on a case-by-case basis, and that's the thing that actually matters.

Ultimately you want criminal justice decisions to be based on reason, and mathematics is the purest form of reason there is. And because you want those decisions to be based on reason, they have to be transparent. Secret methods for arriving at decision-making are fundamentally antithetical to our concept of justice.

Comment Re:"Headlines no more accurate than stupid clickba (Score 2) 161

Alternatively: vendor oversells effectiveness of its proprietary, secret sauce methodology and doesn't like any independent evaluation of its products unless it's favorable. Customers, having a naive faith in technology, buy anyways, which produces exactly the results you mention: programs will be forever terrible at this task. Why should anyone bother to make a program good when customers will shell out good money for mediocre?

Comment Re:No bad software (Score 1) 161

Well, there's bad (i.e., stupid) clients too. They're responsible for a lot of bad software.

If a customer wants to buy magic software without an understanding of what it does or proof that it even works, what are the programmers supposed to do about that? They just report to work and build what their boss tells them to build, and he tells them to build what the customer will buy.

Comment Re: Bankrupt leftist states lead the charge! (Score 1) 353

they never were.
  blue states generally contribute more federal tax dollars than they receive back.
red states generally receive more federal tax dollars than they contribute.
and this while blue states also generally tax themselves higher, and have stronger economies, and better supported lower and middle class populations.

its almost like the GOP mantra of low low taxes doesn't actually help the majority of citizens.

Comment Re:Normally I'm quite against biofuels (Score 1) 261

Look at the price of even a small nuclear power plant. Now look at the price on a Maersk Triple E. There's your answer.

It's just not economically justifiable. So far only one country (Russia) has even found it economically justifiable for icebreakers, which are about the most energy intensive task at sea you can get. As a general rule, reactors only go on ships when they must (when you need them to be deployed for long periods at a time - aka carriers, missile subs, etc)

Comment Re:Normally I'm quite against biofuels (Score 3, Interesting) 261

That would be great if the industry could react that fast, but it takes a lot longer than just a few years to convert a large portion of the world's total petroleum consumption from high sulfur sources to low sulfur sources. They're working on it, but there will be a supply-demand imbalance, and it will have financial consequences.

BTW, the IMO regulations come into effect at the start of 2020, not the end. Not much time left. The rule change was only announced this fall

It can also be dealt with, mind you, by installing scrubbers on ships - then they can still burn high sulfur fuel. But about 80% of shipping is expected to switch to lower sulfur crude, as the capital costs for ships to add scrubbers are quite high. There's another problem, in that the most affordable way to scrub sulfur from ship exhaust ends up dumping it into the sea... but then they're exposing themselves to a liability that years from now that might be banned and they'd have to undergo yet another retrofit.

Slashdot Top Deals

The only thing worse than X Windows: (X Windows) - X