Follow Slashdot stories on Twitter


Forgot your password?
DEAL: For $25 - Add A Second Phone Number To Your Smartphone for life! Use promo code SLASHDOT25. Also, Slashdot's Facebook page has a chat bot now. Message it for stories and more. Check out the new SourceForge HTML5 internet speed test! ×

Comment Re:AI is just software (Score 1) 180

It doesn't guarantee bad things won't happen anymore than following good engineering practices guarantees that a building won't fall over. What it does do is guarantee the ability to trace back to a root cause of why the bad thing happened and pin the responsibility on the appropriate party. As a result, developers have motivation to make sure they're not taking shortcuts, and additionally have ammunition to push back on management if they're ordered to take shortcuts or ignore potential issues. From the management side, management are the ones who have to ultimately sign off on the fact that they were appraised of the risks and deem them acceptable, so they have strong motivation to listen to their developers and make sure that good standards and practices are being followed.

Comment Re:AI is just software (Score 1) 180

So do you think that the methodology laid out for safety critical development would work for AI development as far as chain of responsibility goes? That was actually one of the questions that came up in the software system safety course I took, and unfortunately never got a very good answer (I don't think the instructors really understood how machine learning works well enough to form a good opinion).

Comment Re:AI is just software (Score 1) 180

How do you come up with requirements for a hazard analysis on a heavy machine that can be anywhere in the world at any time, driving at any speed? Your set of conditions that the vehicle will encounter are almost limitless.

You can still do it. From the requirements side, define some reasonable operating conditions and the behavior if it detects itself leaving those conditions. From the safety analysis side, there are multiple methods that are usually used in concert. Generally it'll start with a top down analysis of the energy sources (fuel, kinetic energy in a big moving vehicle, batteries etc.) and work your way down to specific and reasonable failure modes. Then there are a variety of other analysis methods to supplement that, e.g. looking at what would happen if some specific individual component failed and propagated up through the software, etc.

Comment Re:AI is just software (Score 1) 180

I'm not sure how familiar you are with safety critical software and systems (you see it all the time in aviation), but there's actually a pretty well defined process for the entire thing. I'll make a really poor attempt at summing it up:

- A hazard analysis is performed on the system by various engineers (and occasionally even a 3rd party is brought in for peer review). There are a multitude of different ways to go about it, but eventually you end up with a long list of ways the product could fail, with a probability and severity assigned to each failure case.
- After this analysis, everyone comes up with ways to mitigate each of the risks. Removing the risk entirely is preferred, followed by passive safety mitigations are preferred, followed by active, followed by monitoring with alarms. Probabilities and severities are updated accordingly.
- Software is then analyzed in a similar way, except that no probability numbers are assigned. Mitigation steps for software range from self checks (a common example might be to read a sensor on a scale of 0-5v, then read a separate sensor using a separate function that measures the same thing but on a 5-0v scale), to having multiple CPUs of different manufacture running the same code in lockstep and checking each other on the fly. What methods are picked will depend on the hazard analysis and what severity has been assigned to each of the risks

Then in order to be safety certified you need to show documentation that all of those previous steps were followed, as well as show a software process in which:
- There's a clear set of requirements that are traceable to the hazard analysis
- Every line of code is traceable back to those requirements
- There's a set of test cases that are traceable back to the lines of code and the appropriate hazard analysis/requirement
- Documentation showing that all of these test cases have been run (sometimes a 3rd party is brought in to verify this)

Then after all of that is finished, the project managers look at the final risk analysis and sign off on it. They're the ones ultimately responsible for if it fails. In the event that it does fail, they have a stack of paperwork about a mile high to go back and trace how the failure occurred (note: this is the opposite of what Toyota had during the whole unintended acceleration thing). The idea is that in the unlikely event that your software fails and kills someone, you can prove in a court of law that appropriate measures were taken to assess and account for any possible risks.

Comment Re:Industrial accident (Score 1) 407

I witnessed this hack multiple times as a youngster on commercial construction sites.

Rather than removing it from the saw, the framers would wire it into the open position because it allowed them to cut measured lengths of lumber much quicker.

Yep. But it's possible to make those things not be so much of a hindrance (I've definitely used some that were so smooth you wouldn't even know they're there). Making it less annoying is one of the best ways to reduce instances of people circumventing it.

Comment Re:Still want self driving cars? (Score 1) 407

I take it you've never worked on safety critical software which had to meet a certain standard of development and testing, e.g. mil 882 or iso 26262? For that stuff you'd have to have willfully malicious management for bad code to slip through in such a way that it could cause issues (and then they'd sure as shit be liable for it since the whole point of those standards is to leave a paper trail).

Comment Re:Industrial accident (Score 1) 407

Bypassing safety for convenience is a common thing unfortunately, but so are badly designed safety systems that fail to meet the regulations.

I feel like if a safety system is commonly bypassed for convenience, it's a badly designed system. Firstly because it's annoying someone enough that they feel it's worth risking life and limb to circumvent, and secondly because it's easily bypassable.

The blade guard on an old circular saw I once had comes to mind as a good example. It was clearly thrown on as an afterthought of the design, and as a result tended to catch on the wood, screw up cuts, and generally just be a horrible nuisance (I've used other saws in which the guard worked a lot smoother). Eventually I just removed the thing altogether and jammed the interlock.

Comment Re:Unplug it first (Score 1) 407

There are tons of factors that could make it all kinds of different peoples responsibilities. That's why generally with lawsuits like this it's common to fire from the hip and name anyone who could possibly be involved, then let the accused fight it out as to whose fault it is.

If people are doing their job, the robot software team should be able to produce documentation showing what safety went into the code (redundant sensors, secondary processor constantly checking the first one, etc.), the robot hardware team should be able to show that their interlocks and failsafes actually work, the factory management should be able to produce documentation showing proper lockout tagout training, the safety admin types hsould be able to produce documentation showing that those procedures were followed, etc. As a result of that, we'll probably see the case dismissed vs a good number of the accused parties and it'll likely come down to just one or two of them left (the ones who didn't have a paper trail to cover their asses) doing a lot of mutual finger pointing.

Comment Re:You know what's better than streaming? (Score 1) 41

I rarely watch a movie more than once but I listen to the same songs rather often. I don't care for concerts, so I'm afraid I cannot relate on that matter.

That's probably why we differ. For me the streaming model is perfect because while I don't necessarily object to listening to the same song multiple times, what I like and don't like will vary depending on situation, mood, and circumstances. Sometimes I just want semi-rhythmic noise to block out other sounds, sometimes I want something that I can zone out to while programming, and sometimes I want to discover new music that sounds like some other song I heard and liked. I've also found that music I liked 10 years ago I can't stand now, and I don't have any particular loyalty to any particular song or band. As such, I'm more than happy to stream music on a temporary basis (disclaimer: I haven't once paid for a streaming service and probably never will). I feel no need to archive it for later use since there's a 50-50 chance I won't enjoy it anymore anyway after a few years.

Comment Re:Tough shit -- welcome to the real world (Score 1) 283

Wait until it's your turn and we turn the cold shoulder to you when you can't afford your medications when you're 65+ and we tell you "well you should have made better business investments to cover for your retirement shouldn't you have?"

tbh I'd be fine with that, unlike a lot of people I'm capable of planning ahead and saving money. In this hypothetical, my complaints would stem from the fact that I'd spent the last 45 years paying to subsidize others who were incapable of planning ahead and was then told that as a reward for my good planning I'm ineligible for any assistance.

Comment Re:I might consider looking again... (Score 1) 41

skip counts being a big one.

I'm really curious about that use case actually. Do you pick stations just based on "bands I liked at some point in time", or do you actually try to build them around a theme?

Mostly asking because I almost never find myself having to skip songs. The only times I ever run up against the skip limit was when I first create a new station and haven't yet dialed it in to play the kind of music I wanted it to play. If the station is playing stuff that fits within the station but isn't what I want to hear at the time, I realize that I'm on the wrong one for my current work and/or mood and just switch stations to something more appropriate.

Comment Re:Yet another science denier (Score 1) 223

How much energy does it take to create that vacuum under 1 atmosphere? Are you attempting to claim that the laws of energy conservation don't exist when it's a rich guys pet project?

You keep saying this, but I don't think you understand the difference between a sudden release of energy in the space of a few microseconds and the gradual release of energy over the course of multiple seconds. Re-pressurizing a long tube from one end can't happen instantly, so it will most definitely not be the same as releasing whatever amount of energy you think is stored in 1atm of pressure differential all at once.

Slashdot Top Deals

"The value of marriage is not that adults produce children, but that children produce adults." -- Peter De Vries