4 ms·

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

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.

gedy 5y ago
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.
- PragmaticPulp 5y ago
  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.
- humanistbot 5y ago
  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.
roflc0ptic 5y ago
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
- PragmaticPulp 5y ago
  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.
  - dahart 5y ago
    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.
agumonkey 5y ago
maybe match the low reliability of classification with a low impact rule, if the system thinks you're drunk you just have to drive slower ?
- lm28469 5y ago
  Yeah make them do 30mph on the highway this will surely reduce the danger /s
  - agumonkey 5y ago
    Honestly, if the sensor suddenly thinks you're drunk in the middle of a highway segment, there's something very wrong since long before.
    - lm28469 5y ago
      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.
      - agumonkey 5y ago
        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)
        lm28469 5y ago
        Anything short of 100% reliability and 0% error will be a pain in the ass to thousands of people worldwide every single day

gedy 5y ago

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.

PragmaticPulp 5y ago
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.
humanistbot 5y ago
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.

roflc0ptic 5y ago

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

PragmaticPulp 5y ago
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.
- dahart 5y ago
  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.

agumonkey 5y ago

maybe match the low reliability of classification with a low impact rule, if the system thinks you're drunk you just have to drive slower ?

lm28469 5y ago
Yeah make them do 30mph on the highway this will surely reduce the danger /s
- agumonkey 5y ago
  Honestly, if the sensor suddenly thinks you're drunk in the middle of a highway segment, there's something very wrong since long before.
  - lm28469 5y ago
    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.
    - agumonkey 5y ago
      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)
      - lm28469 5y ago
        Anything short of 100% reliability and 0% error will be a pain in the ass to thousands of people worldwide every single day