This is terrible overreach and a good example of magical thinking about the abilities of AI.
Even if the false positive rate of such a system was only 0.1% per driving session, an extraordinarily ambitious goal for any AI system, that still translates to an incredible number of false positives across all drivers. Someone who drives twice a day for work would be expected to hit a false positive every couple of years or so. Do we really want our cars randomly flagging everyone as drunk every once in a while?
Obviously the people writing this bill assumed that perfectly correct AI could exist. They did their job, now the problem is punted to the engineers.
Ridiculous. I’m sure this will get removed eventually, perhaps with further legislation.
I think it’s off base to pin this on AI enthusiasm, that’s just kind of unclear writing in the article. The legislation doesn’t specify a particular anti drunk driving provision, just that the NHSTA study the issue and determine the best course of action. Mandatory breathalyzers seems like what, um, sober minds will come up with. Which certainly feels like stunning overreach to me, but isn’t strictly about AI
Technically the wording is vague enough that they could be referring to anything, but the article specifically highlights distracted driving systems that scan drivers’ eyes:
> Among the systems that have been studied are ones that monitor a driver for signs of distracted, impaired or fatigued driving. One uses sensors that scan drivers’ eyes for signs that are similar to ones that police officers look for when they suspect impairment during traffic stops.
Even breathalyzers aren’t as fool-proof as they sound. Apparently eating certain meals right before using a breathalyzer can trigger false positives. Even breathalyzers have not insignificant numbers of false positives they would become a huge problem if forced upon the entire driving population.
So even with breathalyzers, it’s still about magical thinking about technology’s ability to monitor people’s behavior.
Indeed the wording is vague, and the response to a positive sensor detection is not specified at all. So maybe it’s premature to worry too much about the amount of damage it might do, since we don’t even know what it will do, and it hasn’t passed yet?
Do you think the goal is worth pursuing? Imagine we could achieve zero false positives, would you be in favor of it then? Would it depend on what the system does with a positive detection? If we could have perfect detection of drink driving, should we? Or, said another way, should it be anyone’s right to choose whether they drive drunk?
It’s worth remembering that it was argued that mandatory seatbelts were also terrible overreach, as well as other mandatory safety measures like airbags, child seats, mandatory motorcycle helmets, etc, despite the fact that the safety results are unequivocally positive. It seems like in the US we tend to have a problem with things that are mandatory, regardless of the benefits or outcomes.
The difference is between this and say mandatory seatbelts is that the car won't refuse to start if it thinks that someone isn't buckled. I would refuse to buy a car that did try that, as that's just asking for inconvenient sensor failures that leave you stranded.
There are enough used cars available today that I'm pretty sure I could just buy used for the rest of my life and not have to worry about that.
I don't think I will have to worry about it though because if such legislation ever passed, it wouldn't be long before someone got stranded in the middle of nowhere and died next to a perfectly functional car that refused to start up, the media would make a big deal about it, and the legislation would be reverted.
I see it as an issue of convenience vs actual benefits. Create a system that can tell if I am drunk with near-perfect accuracy from just me being in a vehicle and I am ok with it.
However the current systems we have imply that I'd have to blow into something every time to start my car, that straw thing would get full of saliva, and on top of that there could be false positives so it's obviously a big no.
I assume you mean mandatory breathalyzers as car equipment, alongside high-visibility vest, documents etc. It is law in some countries - I know of France. I judge it one of the most absurd conceivable piece of legislation.
You cannot go to people who are possibly abstinents and suggest that they may drink: it is offensive. Even if well below the median, even if only the third, tenth standard deviation refrained from intoxication, you still cannot go to some random individual and assume they drink - in their culture, that could be a disrespectful to defamatory act.
Plus, it is a matter of prudence: also on a matter of proper relation, you do not want to put those people in the condition of discussing why they find it offensive - forcing them to possibly offend the inquirer.
I think a lot of the proponents for these sorts of rules/restrictions tend to be anti-car and/or throwing bones to their various transport unions, and don't care that they are overreaches or impractical.
The article notes MADD (Mother’s Against Drunk Driving) as one of primary sponsors.
MADD is notorious for demanding government overreach in reducing or eliminating alcohol consumption altogether. They have a history of ignoring any possible negative consequences of their proposals for other people, and this is a good example.
Nope. As the article mentions, the major push is from modern-day temperance organizations like Mothers Against Drunk Driving, which was the major lobby that got the US drinking age raised from 18 to 21. Although you are right that these organizations have authoritarian impulses and don't care that these are overreaches and impractical.
idk man, Discord thinks a picture of my fingers is explicit content, google Lens thinks my gf's cat is a dog, tesla can't detect people _asleep_ at the wheel...
AI is fun until you start applying it to real world issues, then it crumbles apart.
those are all terrible implementations though, nobody would ask a discord like dev team to work on safety critical systems, and musk has a serious problems overselling his driving aid..
doesn't mean a good team of serious guys could make a useful device (I'm no proponent of it btw)
Honda Insights have a "gas cap sensor" that lights the "check engine" light when the gas cap isn't tight enough. "to prevent VOC emissions in California" I'm told. But after nearly 20 years, the sensor isn't reliable anymore, and is hard to find and expensive, so the practical thing is to dike it out.
I've no idea how the "drunk driver" detection is supposed to work; but how is it going to be any more reliable or less troublesome than the gas cap sensor?
Does anyone even in the offices of anyone voting for this crap have the slightest idea what technology they're mandating or how reliable it is? Is there any understanding or concern for people with "auto brewery syndrome" or any of the other edge cases that will trip some folks up?
Or is this purely kabuki "see us care" and damn the cost?
The tech industry has spent decades hyping up AI, convincing the public that they have created human-equivalent general-purpose AI because your phone can do good speech recognition that can read answers parsed from Wikipedia or control your house's lights. The tech press goes along with it, painting a vision that AI can do anything. And then we blame the politicians for buying into the hype? Although I do blame the Republicans for defunding Congress's Office of Technology Assessment in 1992, which offered independent, non-partisan, evidence-based assessments of the feasibility of exactly this kind of law.
Basically anything we can do to make cars more expensive and take them off the road I support at this point, driving in America right now is a frustratingly brutal experience that leaves thousands dead, maimed, late or in prison, every day as well driving the environment into a state of global chaos.
I love these kind of laws, oh we mildly inconvenienced a car company, big deal, they'll end up making more money off it in the end. Hopefully I'll just stop seeing the bodies of people being decapitated or torn in half on my commute to DC and actually be able to actually go where I want on time.
Around 40,000 American die each year on our roads, this has been pretty consistent for decades.
As we add more safety features, the marginal cost per life saved is going up. We got a big bang for the seat belt. But backup cameras cost around $5.7 million dollars in cost for every one life saved.
I commend NHTSA for their hard work, but I can't help but feel government safety mandates in cars are now scraping the bottom of the barrel in cost-effective life saving measures.
If I had to offer an alternative, SAMHSA, the government org focused on substance abuse seems to be under-funded and highlights numerous ways additional dollars can save lives.
So rather than scrape the bottom of barrel on making cars safer, perhaps we can instead spend that money on substance abuse interventions with youth and college age students that lead to alcoholism, binge drinking, and drunk driving. Maybe we would get more bang for our buck.
Unless this part of the bill isn't actually about saving lives, but then I'd have to ask why a Representative from Michigan would have an incentive to make cars more expensive other can wanting to save lives.
Backup cameras probably saved the economy from an insane amount of damages. I've never had an accident at speed, but I do have one insurance claim from reversing into a vehicle (sadly totaled another car because I had a Blazer with brutal rear visibility).
My point was about government mandated safety features.
Backup cameras have value for the consumer in preventing accidents and damage, the government didn't need to mandate their inclusions on cars as consumers voluntarily were willing to add this feature for years.
However, when the government forced all consumers on all to have to add this feature that is where the line was crossed.
I drive a fuel efficient Chevy Spark, it cost me maybe $11000 brand new. I drive an average of 5000 miles a year. When I have to replace the car, it will cost a lot more because now it will have cameras and drunk driving gadgets, etc. I dont even have an automatic transmission or power locks - these features are a waste of money for me, but the government is forcing me to purchase safety features and the economic value to society is spurious. I'd rather there be a sticker where my rear view camera is saying 'in lieu of camera we made a $2000 gift to clean water in Africa, it saved more lives, trust us'
> in lieu of a camera we made a $2000 gift to clean water in Africa, it saved more lives, trust us
Wait, so do you mean you don’t care about the extra cost at all, and you don’t care if Chevrolet charged you a lot more than the cost of a backup cam in order to donate? It’s just the government safety rule you object to?
Or are you saying you think backup cameras cost $2000?
And why didn’t you acknowledge the valid point that the economic value to society of backup cameras is in reduction in accidents, and not limited to loss of life situations?
Are you being honest with a dollars per life calculation? It seems misleading. What about the property damage value, how much is that? What about injuries?
According to the stats I found with a quick search that were used to justify the backup cam rules, there were 15,000 injuries per year in addition to the ~200 deaths. And one third of all backup related deaths are children under 5. What is the economic value of a child, and are you sure it’s less than 5.7 million?
I checked out the part out that mentions backup cameras, and it’s got misinformation contradicted by the report I already posted above. For example, the people talking claim nobody tried to analyze the benefit to people other than children, which isn’t true, and they claim the benefits of backup cams “didn’t work out”, while the report above shows the calculation isn’t expected to pay off until 2028 and isn’t expected to reach maximum benefit until around 2050, but will on the whole be a positive to the economy. “Not that many lives saved” just seems like such an awful position to take, when we’re taking about per-vehicle costs that are less than a week’s gasoline.
Interesting, thanks. I think the transcript was specifically calling out the low value of human lives that were used in the past. This report seems a lot more rigorous than the other investigations and articles I've seen.
This is also from 2014, which means the net cost/vehicle has probably come down a fair bit (camera modules are <$5 and most cars including entry-level are shipping with screens already).
Sometimes you need government mandates otherwise the industry standardizes a model that actually costs the consumer more. Bumper height and construction for example should be standardized because the incentive is for ineffective bumpers.
Forward visibility and pedestrian impact standards need to be set to prevent monster grills. A lot of this will actually lower total cost of ownership at little or no cost.
I found a stat that suggested ~100 [1] (another stat says 500 and 15,000 injuries (!) [2] and there's a 3rd between them [3]), so say 600 million. Given the insurance cost of replacing vehicles, that would only be 30k accidents prevented to pay for itself.
FWIW, I've heard anecdotally that the insurance companies pushed for backup cameras more than anyone else, but given the benefits to lower insurance (arguable, but in theory), less deaths and less injuries, it seems like a win. The technology will also get cheaper over the years.
I’ll take the backup camera even if it costs more than my expected savings on car damages because of the value of not having your day ruined by needing to talk to insurers. I got yelled at once by locals when I slightly nudged a car behind because they assumed I wouldn’t tell the owner of the other car! That sort of shit as well!
America has an abysmal annual death rate for automobiles (per one million citizens) compared to the Nordic countries (104 vs 27 in 2013). US is one of the worst in the developed world.
Clearly, something is being done correctly everywhere else.
> Around 40,000 American die each year on our roads, this has been pretty consistent for decades.
Isn’t this a misleading framing? It suggests that safety isn’t improving, but the population has increased, as have speed limits and road miles traveled. The per-capita death rates have been falling consistently for decades. See the first graph at the top of this page: https://en.wikipedia.org/wiki/Transportation_safety_in_the_U...
>If I had to offer an alternative, SAMHSA, the government org focused on substance abuse seems to be under-funded and highlights numerous ways additional dollars can save lives.
>So rather than scrape the bottom of barrel on making cars safer, perhaps we can instead spend that money on substance abuse interventions with youth and college age students that lead to alcoholism, binge drinking, and drunk driving. Maybe we would get more bang for our buck.
Keep in mind though, it isn't tax money that's going to fund this, it's consumers, who will pay in the form of increased car prices. This means disenfranchising poorer people from being able to afford to own cars and/or reducing other features, such as increased passenger safety and fuel efficiency.
> We got a big bang for the seat belt. But backup cameras cost around $5.7 million dollars in cost for every one life saved.
Where does that figure come from?
Car sellers don't sell OEM cameras at cost. They have a ridiculous markup (between 5x & 10x and yes that includes certification & testing). Your $500 dollar reverse cam is cost at ~$50.
By mandating them, the 'luxury!' margin will magically disappear.
This sounds less like an anti-drunk driving measure and more of a permanent general driving ability monitor.
Surely if they wanted to target alcohol specifically they would just mandate breathalyser to start the engine, which is already a thing in many countries for people previously convicted of DUI
Look, I know that drunk driving is a serious issue, but based on what I see on the roadway, NOBODY is paying any damn attention out here anymore. Whether it is phones or giant screened infotainment systems, the amount of distractions is staggering. If we are going to go the route of overreach and “nanny state”, why not find a way to disable certain features of the drivers phone and do away with the large screen infotainment systems altogether.
Used to commute on one of those double decker tech busses down a few Interstate highways in the Bay Area, and from those you can get a good view into the cabins of other commuters’ cars. Every day you can see about 50% of drivers playing with their phones while driving. And it’s not just people occasionally glancing at their phones in their cup holder. They are holding the phone in their hand, scrolling through god-knows-what, typing/texting, watching videos. It’s totally insane! At least half of drivers are doing this. What can possibly be this important on Instagram that you’re willing to risk your life for?
If I could get 10% of the fine, I could quit my job, stand by the road and record all the drivers who did that. I would make bank.
Around here the fee for talking with a handheld phone is more than twice what a headset costs + points off your license. Yet people still do it all the time.
According to [1], in California, total out of pocket for a first time cell phone violation is ~$150, total out of pocket for a subsequent offense is ~$250. Let's conservatively assume that for 75% of drivers, it's a first time violation. Average fine is $175.
According to random internet search [2], the saturation flow of a lane of traffic is about 1900 vehicles per hour. Let's assume a 4 lane highway at rush hour is about 80% of saturation, which yields 6,000 vehicles per hour.
Assume 50% are offending, and you can positively catch/convict 10% of those, so 300 vehicles x $175 = $262,500 in fines per hour. At 10% finder's fee, you could make $2,625 PER HOUR just sitting by the side of the road with a camera. Not bad!
I was in a wreck 2 months ago where I'm assuming the woman was on her phone, because she slammed into me while I was stopped at an intersection. I am still arguing with her insurance company over the property damage claim, and we haven't even started on the bodily injury claim because I haven't yet finished all my health-related crap caused by the "accident".
The phone companies, insurance companies, car companies, and individuals have no incentive to disable phones while driving. Who is left to fix this problem?
Maybe it will get fixed when a senator's family member is killed by someone driving and using their phone.
"Safety" has become an euphemism for "control". It's as if politicians speak a coded language: one of them says "we want safer cars" and others translate that to "we want more control over cars". If someone says "to protect children we need these measures" others understand that it really means "we need these authoritarian measures".
> It orders NHTSA to study the feasibility of various technologies and establish a final rule within three years mandating some form of anti-drunk driving technology.
This kind of administrative rulemaking is required to go through multiple rounds of public notice and comment. The EFF & related orgs should be intervening and lobbying at every stage in this three-year long process.
Even if one agrees these are good things and needed, why does our congress not pass these laws individually so they can be voted and understood for their own worth?
Lots of reasons, one of which is that the Senate is rigged to favor rural places and doesn't represent the majority of Americans anymore. Even reasonable, popular bills are unable to pass.
The Senate was rigged from the jump to favor states with lower populations. That clause is also the only part that is immune from the amendment process. It's never represented the people, it's always represented the states.
The problem at the moment is that the person in charge of one of the two equally sized factions in power has zero interest in doing any kind of leadership or legislating. That's why nothing is happening.
> It's never represented the people, it's always represented the states.
...which is a lot less of a problem if the House and the Presidency represent the majority. Then the Senate functions as intended: a check on the "mob" of democracy.
But now we're in a position where most states are gerrymandered, the Electoral College also heavily favors the rural minority, and the judiciary has been intentionally hijacked by abusing the rural bias.
It makes sense if you think about it. Given hundreds of years to figure it out, eventually a minority interest is going to find a way to engineer the federal system to seize (and retain) power against the will of the majority.
Packaging things this way lets representatives and senators more effectively barter for support - "I'll support this bill if you add this other thing" is more effective tactic with these large omnibus bills.
You can argue if this is good or not for the country, but the argument definitely is that it allows a representative more tools/leverage for his constituency, and that it allows for more effective compromises.
Any attempt to reversing this trend requires dealing with that line of argument.
No. There's a pretty big difference between “you are not allowed to do <thing that humans naturally do all the time>” and “you should be aware that <thing that humans naturally do all the time> is a potential source of error rather than a solid step in reasoning.”
Point out that something characterized in the former manner is, in fact, the latter is not “a nitpick”.
As someone who has convicted of DUI twice in my youth, I'm 100% in support of keeping people off the road if they're drunk. The two times I got caught are just that: the times I was caught. I deserved many many more. We can't stop people from thinking it's okay to drive when they're drunk because they're drunk. Driving is not a right. Keep people like my younger self from making potentially deadly decision.
It seems like insurance companies should be liable for massive damages for DUIs (raising rates), but offer the option to "voluntarily" install a breathalyzer to get lower rates. That feels like the American way.
Perfectly pricing risk implies omniscience (ie totalitarianism), which means that pricing risk is not an unquestionably good thing. Furthermore, markets can never be fully efficient, as that would imply P == NP. And so what is actuarially considered and ignored is itself a public policy choice. Stating things in the passive terminology of "the market" is just obfuscating that choice.
This bill keeps getting worse with all the wreckage that is getting included. Please please, start including the names of the reps pushing these idiotic measures as it's the only way to stop them.
And before someone accuses me of "not wanting to stop drunk driving", of course I do. The way to do that is to address the underlying causes of alcoholism, not patching the symptoms.
You can address causes til you're blue in the face. We've already tried a dozen ways - state-run liquor stores, therapy programs, breath testing, prohibition and on and on.
It may be time for something different. Maybe just stop a car cold that's behaving erratically. Nobody has the right to endanger strangers on the highway, for any reason.
The right bill will be important. But its not idiotic, to try to save lives with technology. Nor is it far away from a workable solution.
Just do away with the legal limit altogether, drop it to zero. If you've had anything at all to drink then you should not be operating a motor vehicle on puclic roadways. If you get caught DUI then your license should be permanently revoked. I wonder how drastically improved the alcohol related traffic fatality statistics would be with that law. The regulation currently up for consideration clearly violates the fourth amendment.
Armenia, Azerbaijan, the Czech Republic, Hungary, Jordan, Kyrgyzstan, all have BAC limits of zero. Try to leave constructive comments that discuss the points made in a parent instead of attacking it outright and calling it names.
If this happens I imagine used cars will be more valuable than new ones (and not because people want to drive drunk but because its impractical and requires more stuff that can fail + higher costs)
Even if the false positive rate of such a system was only 0.1% per driving session, an extraordinarily ambitious goal for any AI system, that still translates to an incredible number of false positives across all drivers. Someone who drives twice a day for work would be expected to hit a false positive every couple of years or so. Do we really want our cars randomly flagging everyone as drunk every once in a while?
Obviously the people writing this bill assumed that perfectly correct AI could exist. They did their job, now the problem is punted to the engineers.
Ridiculous. I’m sure this will get removed eventually, perhaps with further legislation.