What do experts think of the future of driverless cars? Any surveys?
January 8, 2018 12:57 PM Subscribe
I'm looking for information on opinions among "experts" on the future of driverless/self-driving/autonomous cars. Do most experts agree we'll have some version of them soon? Is there a lot of disagreement?
The ideal thing I'm hoping to find is: out is a recent survey that shows: What percentage of experts think we'll have, say, x% of level Y cars in the road by year YYYY.
But I'd also settle for something more anecdotal that says "everyone who works in the field pretty much agrees it's pretty likely we're on the path to ____", or "there's a lot of disagreement in the field about _____."
This is a for help with a conversation with a friend who is particularly suspicious of what he perceives as corporate hype. So I'm hoping to find data on opinions among non-corporate experts: Academic researchers primarily, but maybe also government policy experts.
The ideal thing I'm hoping to find is: out is a recent survey that shows: What percentage of experts think we'll have, say, x% of level Y cars in the road by year YYYY.
But I'd also settle for something more anecdotal that says "everyone who works in the field pretty much agrees it's pretty likely we're on the path to ____", or "there's a lot of disagreement in the field about _____."
This is a for help with a conversation with a friend who is particularly suspicious of what he perceives as corporate hype. So I'm hoping to find data on opinions among non-corporate experts: Academic researchers primarily, but maybe also government policy experts.
I work in the transportation planning field, and boy howdy is there ever a lack of a consensus on what exactly the future holds regarding autonomous vehicles. I think that Todd Litman's summary [PDF] is a good look at a number of the issues - he runs his own little transportation analysis firm, is definitely not corporately influenced, and does some very good summaries of key issues. This article by him covers most of the bases in a shorter form with less footnotes.
My own two cents quickly -- I think that a more gradual timeline for adoption, particularly at the "every car is autonomous" level is likelt. AVs are not being sold yet and there can be hurdles in terms of development even when a technology is very close to being ready. From the date"true" AV technology is made public, there will be a substantial delay until all new cars are AVs.
Car manufacturers introduce new features on luxury vehicles to justify the price premiums, and gradually migrate them down through the lines until they become standard on all models. For instance, backup cameras will become mandatory on all new vehicles starting this year, 18 years after they were first introduced. Airbags took 15 years, automatic transmissions have been widely available for 70 years and they still aren't everywhere. It will likely be roughly 20 years before autonomous vehicles go from being a premium option on Lexuses to being the standard on cheap compacts.
Cars are also durable objects; few vehicles spend less than 10 years on the road (barring write-off incidents), and there is a long tail of vehicles - looking at one survey of a Canadian city, 5% of the vehicles are over 20 years old, and almost 2% are over 30 years old. In other words, airbags were invented in 1951, entered the market in 1974, were mandatory in (driver's side) in all new vehicles by 1989, and still 2% of the cars on the road today were manufactured before airbags were mandatory.
posted by Homeboy Trouble at 1:40 PM on January 8, 2018 [10 favorites]
My own two cents quickly -- I think that a more gradual timeline for adoption, particularly at the "every car is autonomous" level is likelt. AVs are not being sold yet and there can be hurdles in terms of development even when a technology is very close to being ready. From the date"true" AV technology is made public, there will be a substantial delay until all new cars are AVs.
Car manufacturers introduce new features on luxury vehicles to justify the price premiums, and gradually migrate them down through the lines until they become standard on all models. For instance, backup cameras will become mandatory on all new vehicles starting this year, 18 years after they were first introduced. Airbags took 15 years, automatic transmissions have been widely available for 70 years and they still aren't everywhere. It will likely be roughly 20 years before autonomous vehicles go from being a premium option on Lexuses to being the standard on cheap compacts.
Cars are also durable objects; few vehicles spend less than 10 years on the road (barring write-off incidents), and there is a long tail of vehicles - looking at one survey of a Canadian city, 5% of the vehicles are over 20 years old, and almost 2% are over 30 years old. In other words, airbags were invented in 1951, entered the market in 1974, were mandatory in (driver's side) in all new vehicles by 1989, and still 2% of the cars on the road today were manufactured before airbags were mandatory.
posted by Homeboy Trouble at 1:40 PM on January 8, 2018 [10 favorites]
Response by poster: Maybe one specific question I'd ask is this: My friend's position is something like "There's no way we'll see self-driving cars on the roads in any significant number (say over 5%) in our lifetime (say- the next 40 years).
My understanding (maybe very incorrect) is that his position falls ouside the range of opinions you'd find among most experts in the field- that when you look at the range of expert opinions out there, the more skeptical foks are saying "we'll have them, but it may take 25 years till they get common".
Buy maybe my understanding is wrong? Maybe I've been swept up in a lot of over-optimistic industry hype?
posted by ManInSuit at 2:30 PM on January 8, 2018
My understanding (maybe very incorrect) is that his position falls ouside the range of opinions you'd find among most experts in the field- that when you look at the range of expert opinions out there, the more skeptical foks are saying "we'll have them, but it may take 25 years till they get common".
Buy maybe my understanding is wrong? Maybe I've been swept up in a lot of over-optimistic industry hype?
posted by ManInSuit at 2:30 PM on January 8, 2018
So, I'm a researcher in this space (at a university, funded by the research arm of a major automaker). The general expectation, from talking to research and development staff at said funder (and others in the last couple years), is that we'll see lower-level automation (e.g., L2 [capable in some areas, with supervision]) in quite a few models in the next few years. Higher-level automation, like L4 (totally capable when engaged, the driver will never need to intervene, but the vehicle may need a driver in some areas, or L5 (capable in any environment) is a ways out. I don't know of any credible claim that sees real L4 on less than a five year timeline.
Wrapped up in all of this is that anything in this space takes an understanding of the driver that we really don't have. Let's say you've got a L2 automated vehicle (like, say, a Tesla with Autopilot). It assumes you're paying attention and able to take over whenever it needs you to... except that drivers who use Autopilot treat it like a L3 system (where the vehicle doesn't expect you to pay attention to the roadway, and just needs to be able to notify you when it needs you to do something). So, I'd say autonomous vehicles are coming, but they're a ways out right now. We'll see L2 autonomy in a lot more vehicles a lot sooner, but we won't be seeing L4 or L5 for a while, and that's probably a good thing.
On preview: ManInSuit I'd say we won't see meaningful numbers of L4 or better autonomous vehicles for a decade, and even once they show up, the transition to a fully-autonomous fleet is expected to be 25 years or more. We'll see vehicles capable of limited autonomy (L2) much sooner than that, and I'd expect them to be the majority of the US fleet within 20-ish years.
posted by Making You Bored For Science at 2:41 PM on January 8, 2018 [2 favorites]
Wrapped up in all of this is that anything in this space takes an understanding of the driver that we really don't have. Let's say you've got a L2 automated vehicle (like, say, a Tesla with Autopilot). It assumes you're paying attention and able to take over whenever it needs you to... except that drivers who use Autopilot treat it like a L3 system (where the vehicle doesn't expect you to pay attention to the roadway, and just needs to be able to notify you when it needs you to do something). So, I'd say autonomous vehicles are coming, but they're a ways out right now. We'll see L2 autonomy in a lot more vehicles a lot sooner, but we won't be seeing L4 or L5 for a while, and that's probably a good thing.
On preview: ManInSuit I'd say we won't see meaningful numbers of L4 or better autonomous vehicles for a decade, and even once they show up, the transition to a fully-autonomous fleet is expected to be 25 years or more. We'll see vehicles capable of limited autonomy (L2) much sooner than that, and I'd expect them to be the majority of the US fleet within 20-ish years.
posted by Making You Bored For Science at 2:41 PM on January 8, 2018 [2 favorites]
There are already (semi) autonomous cars on the roads (self-ish link to a MetaFilter post).
National Highway Traffic Safety Administration (NHSTA) currently lists "five eras of safety" that offers one non-commercial/governmental forecast, which currently foresees the next two eras being:
2016 - 2025
Partially Automated Safety Features (existing range of options)
- Lane keeping assist (exists now)
- Adaptive cruise control (exists now)
- Traffic jam assist (exists in various forms now)
- Self-park (exists now)
2025+
Fully Automated Safety Features
- Highway autopilot (currently exists, with caveats)
Mind you, there are levels of automation, as noted by Making You Bored For Science. Society of Automotive Engineers (SAE) has 0 to 5, and that's the US standard that is constantly being updated. NHSTA doesn't clearly spell out how their forecast relates to the levels (NHSTA uses SAE's scale), so that gives them even more leeway to say "look, we were right!"
In short, most professionals agree that semi-autonomous vehicles are near-term, but full automation (no steering wheel) is a ways off yet. It's a case incremental progress.
posted by filthy light thief at 2:51 PM on January 8, 2018 [2 favorites]
National Highway Traffic Safety Administration (NHSTA) currently lists "five eras of safety" that offers one non-commercial/governmental forecast, which currently foresees the next two eras being:
2016 - 2025
Partially Automated Safety Features (existing range of options)
- Lane keeping assist (exists now)
- Adaptive cruise control (exists now)
- Traffic jam assist (exists in various forms now)
- Self-park (exists now)
2025+
Fully Automated Safety Features
- Highway autopilot (currently exists, with caveats)
Mind you, there are levels of automation, as noted by Making You Bored For Science. Society of Automotive Engineers (SAE) has 0 to 5, and that's the US standard that is constantly being updated. NHSTA doesn't clearly spell out how their forecast relates to the levels (NHSTA uses SAE's scale), so that gives them even more leeway to say "look, we were right!"
In short, most professionals agree that semi-autonomous vehicles are near-term, but full automation (no steering wheel) is a ways off yet. It's a case incremental progress.
posted by filthy light thief at 2:51 PM on January 8, 2018 [2 favorites]
Mod note: A couple deleted. Please note that OP is specifically asking for expert opinions.
posted by taz (staff) at 6:30 AM on January 9, 2018
posted by taz (staff) at 6:30 AM on January 9, 2018
« Older Should I try to contact my ex-best friend again? | Online dating, I ask him for coffee, he agrees... Newer »
This thread is closed to new comments.
(And links to a very smart non-survey article on the subject.)
posted by Smearcase at 1:39 PM on January 8, 2018