Senator asks Tesla to disable AP until they fix "anti-nag" tricks.

[Daniel in SD]

Why would it respond differently? It is effectively just a mass attached at an attachment point. Yes, there's some amount of elasticity that varies according to how much the person tenses up, but the details are person-dependent, so I can't see any reason you couldn't some up with something that would have a plausible amount of elasticity, and the harder you try to build software to detect such a defeat device, the more real people will get nagged for no reason, which drives down adoption and use of autopilot, which reduces safety far more than any (mostly theoretical) gains from preventing people from driving with hands off while paying attention.

Because an arm has multiple pivot points and is attached to a body.

 

[bhzmark]

Mass residents should communicate something like this to Markey and his staff:


Why are you against technology that is proven to reduce collision frequency?

"When we evaluated Teslas with the version 1 hardware after the Autopilot software was deployed, we saw a significant reduction in collision claim frequency"

Fewer physical damage, injury liability claims for Model S with advanced features

 

[Daniel in SD]

Why are you against technology that is proven to reduce collision frequency?

"In this limited analysis, HLDI found that the frequency of claims filed under PDL, BI, MedPay and PIP didn't change once Autopilot was enabled, but the frequency of collision claims fell by 13 percent."
Hardly conclusive! It also seems possible that the other safety features were updated and improved at the same time as the Autopilot release. Using data from right after Autopilot was released also seems problematic. Tesla says that most Autopilot accidents occur among very experienced users.
Tesla's data doesn't seem to be enough to make any conclusions because they don't correct for when and where people use Autopilot.

 

[NoMoGas]

Sorry, he's a putz who has been going after Tesla for some time now... It's why I'm supporting Kennedy.

 

[Ludalicious]

Sen. Markey sent a letter to Tesla asking them to disable AP until they fix the anti nag tricks. You can watch it here:

Tesla is under pressure by Senator over Autopilot claims and nag defeat tricks - Electrek

I definitely think he has a point about the anti-nag tricks being bad and dangerous. I believe the autopilot buddy device is already illegal.

Ideally, Tesla should use a camera based driver monitoring system but unfortunately, that is not going to happen any time soon plus it could not be retrofitted into existing cars.

So what can Tesla do to solve this problem? Is there a software fix? Could Tesla modulate the torque requirement slightly so that a constant torque would not work anymore?

How exactly do you expect to use a rear camera on vehicles not equipped... How about people dont be an idiot behind the wheel.

 

[dgatwood]

Because an arm has multiple pivot points and is attached to a body.

I never got around to replying to this, but I might as well.

If you were detecting it in three dimensions, you'd be able to tell the difference trivially. But you're talking about detecting the difference with a steering wheel that only moves in a fixed arc, without any capacitance sensing to detect where on the wheel the person's hand likely is. It's much, much harder to make sense of such limited data.

As it stands, the Tesla torque sensors can't reliably tell the difference between me with two hands on the wheel and me with zero hands on the wheel. In theory. that ought to be about a million times easier than what the Senator is asking for. And if they don't even have precise enough measurements to pull that off, they have no prayer of usefully detecting anti-nag hardware.

And I don't believe for one minute that it would be possible to reliably do so without incurring a false warning rate high enough to discourage autopilot use, which would be counterproductive from a safety perspective.

I'd bet on FSD shipping first. Just saying.

 

[Xenoilphobe]

Interesting that $enator Markey is the top recipient of PAC donations from the Transportation Unions, Toyota, and the United Autoworkers Union and he loves taking money from lobbyists ~$750,000 just from them alone.

Sounds like you can't buy this guy, you can only rent him.

Sen. Ed Markey - Massachusetts

• Summary
• Elections
• Industries
• PACs
• Contributors
• Geography
• Demographics
• Expenditures
• Legislation
• News
• Other Data

Member
Candidate
Zip
Contributors 2015 - 2020
SELECT A CYCLE
2020

SELECT A PROFILE
Campaign Committee Only
Leadership PAC Profile Only
Campaign Committee & Leadership PAC Combined
$2,001,014

grand total of contributions Ed Markey has reported in the current election cycle.
Number of Contributions from Individuals (of $200 or more): 1,694

Top 20 contributors to Campaign Committee
Download .csv file
Top:
20
Rank Contributor Total Individuals PACs
1 WilmerHale Llp $66,439 $66,439 $0
2 Akin, Gump et al $33,899 $33,149 $750
3 Granite Telecommunications $33,400 $33,400 $0
4 Democracy Engine $31,050 $31,050 $0
5 Blackstone Group $30,000 $30,000 $0
6 Bain Capital $27,039 $27,039 $0
7 DLA Piper $25,900 $15,900 $10,000
8 Robbins, Geller et al $25,250 $25,250 $0
9 DISH Network $24,200 $14,200 $10,000
10 Mintz, Levin et al $22,965 $22,965 $0
11 Rock Creek Group $22,000 $22,000 $0
12 Harvard University $21,036 $21,036 $0
13 Beacon Capital Partners $18,900 $18,900 $0
14 Northwest Bio Therapeutics $16,700 $16,700 $0
14 Third Rock Ventures $16,700 $16,700 $0
16 Artery Capital Group $16,600 $16,600 $0
16 Continental Wingate $16,600 $16,600 $0
18 Walmart Inc $16,220 $11,220 $5,000
19 Nixon Peabody LLP $15,730 $12,130 $3,600
20 Brownstein, Hyatt et al $15,500 $13,500 $2,000
*registrants, or active lobbying firm

These tables list the top donors to candidates in the 2015 - 2020 election cycle. The organizations themselves did not donate, rather the money came from the organizations' PACs, their individual members or employees or owners, and those individuals' immediate families. Organization totals include subsidiaries and affiliates.

Why (and How) We Use Donors' Employer/Occupation Information

METHODOLOGY

NOTE: All the numbers on this page are for the 2015 - 2020 election cycle and based on Federal Election Commission data released electronically on December 16, 2019. ("Help! The numbers don't add up...")

Feel free to distribute or cite this material, but please credit the Center for Responsive Politics. For permission to reprint for commercial uses, such as textbooks, contact the Center: [email protected]

 

[bhzmark]

"In this limited analysis, HLDI found that the frequency of claims filed under PDL, BI, MedPay and PIP didn't change once Autopilot was enabled, but the frequency of collision claims fell by 13 percent."
Hardly conclusive! It also seems possible that the other safety features were updated and improved at the same time as the Autopilot release. Using data from right after Autopilot was released also seems problematic. Tesla says that most Autopilot accidents occur among very experienced users.
Tesla's data doesn't seem to be enough to make any conclusions because they don't correct for when and where people use Autopilot.

Its data that speaks for itself. Do you have any data (not anecdotes) supporting that cars with AP available or its use is less safe in the aggregate?

Or otherwise what is your point? Just complaining that more data would be helpful? Sure. But meanwhile I think this data from third party insurance companies should be even more reliable.

 

[Daniel in SD]

Its data that speaks for itself. Do you have any data (not anecdotes) supporting that cars with AP available or its use is less safe in the aggregate?

Or otherwise what is your point? Just complaining that more data would be helpful? Sure. But meanwhile I think this data from third party insurance companies should be even more reliable.

My complaint is that we don't have the data to make any conclusions either way.
Data definitely does not speak for itself. If the data says that people get in more accidents when their headlights are on does that mean we should remove headlights from cars?

 

[bhzmark]

My complaint is that we don't have the data to make any conclusions either way.

Outside of RCTs, there is always possible confounding variables that cannot be perfectly controlled, but intelligent people try to identify them. Here they specifically separated it out from other tech changes and found a significant difference. And IIHS are in the business of exactly this. And it is the best data we have.

It could have showed MORE collision claims, but it didn’t, it showed LESS. What else can account for the difference? Maybe the people who bought AP are better drivers anyway. You can have an intelligent discussion about whether the difference is better explained by something else. But the data is simply there and there isn’t any better variable to explain the data.

 

[timewasted]

The product managers that I work with constantly say that given some set of data, they can make it tell whatever story they need to tell at the time. They’re not fudging the data, they’re just picking the most favorable way to interpret the data.

 

[cucubits]

Data definitely does not speak for itself

 

[willow_hiller]

Just as a follow up to this thread, I thought I'd share a surpisingly written article on the subject by Motortrend: https://www.motortrend.com/news/turning-autopilots-autonomous-driving-instructors/

Long story short, we need to take the personal responsibility to treat Autopilot with the seriousness a beta ADAS needs to be treated.