In tech, we fear what we can’t control

That is the topic of my new Bloomberg column, here is one bit:

Like drones, driverless cars possess some features of an especially potent scare story. They are a new and exciting technology, and so stories about them get a lot of clicks. We don’t actually know how safe they are, and that uncertainty will spook people above and beyond whatever is the particular level of risk. Most of all, driverless cars by definition involve humans not feeling in direct control. It resembles how a lot of people feel in greater danger when flying than driving a car, even though flying is usually safer. Driverless cars raise a lot of questions about driver control: Should you be allowed to sleep in the backseat? Or must you stay by the wheel? That focuses our minds and feelings on the issue of control all the more.

And:

The recent brouhaha over Facebook and Cambridge Analytica (read here and here) reflects some similar issues. Could most Americans clearly and correctly articulate exactly what went wrong in this episode? Probably not, but people do know that when it comes to social networks, their personal data and algorithms, they don’t exactly feel in control. The murkiness of the events and legal obligations is in fact part of the problem.

When I see a new story or criticism about the tech world, I no longer ask whether the tech companies poll as being popular (they do). I instead wonder whether voters feel in control in a world with North Korean nuclear weapons, an erratic American president and algorithms everywhere. They don’t. Haven’t you wondered why articles about robots putting us all out of work are so popular during a time of full employment?

We are about to enter a new meta-narrative for American society, which I call “re-establishing the feeling of control.” Unfortunately, when you pursue the feeling rather than the actual control, you often end up with neither.

Do read the whole thing.

Comments

yes good post. Children worry a lot about quicksand, even though it doesn't tend to be a problem in adult life.

I for one still worry about quicksand.

Comments for this post are closed

People do no sink in quicksand. Fluidized sand is much denser than water and if you go into a fluidized sand biofilter, you float like a cork and will need up to 75 lb of lead weights to dive to the bottom for inspections.

A fluidized peat bog will allow you to sink with enough fiber to prevent efficient swimming.

Quicksand may not be real but I have found myself is sucking high smectite mud stuck up to my waist enough times to know it needs a name of its own.

Comments for this post are closed

Comments for this post are closed

Another childhood myth exploded, like UFOs and Bigfoot. :^(

Comments for this post are closed

Comments for this post are closed

The recent brouhaha over Facebook and Cambridge Analytica (read here and here) reflects some similar issues. Could most Americans clearly and correctly articulate exactly what went wrong in this episode?

That would be easy - Trump won and it is driving the Left nuts. They have to find some external reason to blame for their failure because they cannot admit that the Deplorables are right and Hillary was an awful candidate. So they are lashing out in irrational ways. They want to find a witch to burn - but they keep coming up short.

It is not *people* who feel a loss of control with Facebook. It is the elites. They control things like CNN and yet Trump used Twitter to talk right over the top of them to the voters. Directly. They are panicking. The solution is to regulate the living hell out of social media so no one else can ever do that again. Shamefully the Super Elite 0.00001% who own these companies are co-operating.

So your answer is no, you don’t understand.

Can I like or upvote James reply?

Comments for this post are closed

Comments for this post are closed

SMFS understands perfectly.

As for the sad, declining status of Tyler Cowen, well....

>an erratic American president

No kidding! Remember that time he smuggled millions and millions of dollars into Iran on pallets, and didn't tell anyone about it? My God, how erratic was that?

Yes, and that erratic president is now living a party life full time, hanging with the super rich on yachts and Martha's Vineyard.

Amazing how those professional SJWs turn into one percenters.

"... you couldn't tell the pigs from the farmers..." - last line in "Animal Farm" by George Orwell.

I don’t what the Obamas’ net worth is but I doubt it will be full on Clintonian in 10 years. Hillary and Bill went from -500,000$ to 240,000,00. Obama was out of his depth, as Trump may prove to be too, but he wasn’t as venal and tone deaf as the Clintons.

Who in your opinion was the last president not to be 'out of his depth'? Agree re relative venality, of course the Obamas are cashing in, it goes with the job (modern era mainly), but the Obamas are total amateurs compared to the Clintons.

Comments for this post are closed

"I don’t what the Obamas’ net worth is but I doubt it will be full on Clintonian in 10 years"

The Obama's obtained an estimated $65 million book advance a year ago.

https://www.vox.com/culture/2017/3/2/14779892/barack-michelle-obama-65-million-book-deal-penguin-random-house

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Related:

"Under Bannon’s leadership in 2014, Cambridge Analytica found that young, white conservatives responded well to “thinly veiled forms of racism against African Americans” and authoritarianism. Strangely, the only foreign issue they tested was views of Putin."

https://t.co/8GqxApol19

But then we knew that, from these pages. Marginal Revolution was a microcosm.

All those comments are from a Canadian programmer, who actually wrote a good chunk of the code. I suspect the items that he classifies as “thinly veiled forms of racism against African Americans” are highly subjective. After all he wouldn't use the qualifier "thinly veiled" if it was actually clear cut racism.

"But then we knew that, from these pages. Marginal Revolution was a microcosm."

What's amazing to me is how you think this matters? Even if hypothetically the Right is more racist in their bigotry versus the Left's cultural bigotry, so what? It's clear the Left is far more open about their bigotry. Just two days ago clockwork_prior was making openly bigoted remarks about "people from Alabama". No one on this board has routinely make similar comments about "blacks from Alabama".

Furthermore, there's plenty of cases of Left wing racism. Indeed, the Left is often public about their racism. We have the overt racism of the Latina group La Raza and the Black Lives Matter movements.

With that hand-waving, you show us who you are, man.

Translated -> "If you point out my bigotry, you must be a nazi".

The new line from the left is that opposition to any of their agenda to enshrine laws and institutions dedicated to identity-based discrimination, you are a Nazi and you deserve to be physically attacked.

Comments for this post are closed

Anyone who as been here for any length of time knows that I have never called for anyone to be "physically attacked."

Maybe you should worry that people will start to see your madness as characteristic of your movement.

Comments for this post are closed

"Anyone who as been here for any length of time knows that I have never called for anyone to be “physically attacked.""

Richard Spencer doesn't have to call for physical violence. He can simply present premises that lead to violence being the logical conclusion, not dissimilar to what you and the broader left do when you claim that Mitt Romney will enslave the black population or anyone who wants to reduce immigration is an actual Nazi or anyone who opposes you handily "[shows] us who you are, man". That implication is a wink and a nod to black-clad Democrats intent on beating up their opposition.

Comments for this post are closed

In related news:

Former VP Joe Biden talking about Trump: ""A guy who ended up becoming our national leader said, 'I can grab a woman anywhere and she likes it,'" Biden said. "They asked me if I’d like to debate this gentleman, and I said 'no.' I said, 'If we were in high school, I’d take him behind the gym and beat the hell out of him.'"

Comments for this post are closed

Comments for this post are closed

JWatts wants us to know that "thinly veiled racism" is out. Full-fledged racism is the new... black!

Comments for this post are closed

Comments for this post are closed

Bringing Putin into the discussion of Facebook is instructive in that they both represent such recent flipflops by the Democrats. In 2012, Facebook was celebrated for its ability to boost the Obama campaign, and Obama was telling Putin that he'd have "more flexibility" after the election. SMFS is right that these aren't existential fears but rather are highly situational political cudgels.

Fool me once, shame on you; fool me twice, shame on me.

The first "reset" was the once. Anyone trusting (or congratulating) Putin now is into high counts, much higher than twice.

I think that's a fair point. However, it's pretty hard to take the Left's point on this too seriously, when their attitude 5 years ago was precisely the opposite of what it is now.

http://foreignpolicy.com/2012/03/09/obama-congratulates-putin-for-election-win/

There's a nice picture of Obama with a smile while he's shaking Putin's hand.

Comments for this post are closed

Are you faulting Obama for learning?

In the present situation that would be Ironic with a capital-I.

Comments for this post are closed

No, I'm faulting Obama for being a dumb ass. Putin was never trustworthy. Obama didn't just shake his hand and congratulate him, he also took Putin's word that Russian wouldn't get involved in Syria. Putin played Obama for a fool.

"The so-called “red line” episode in September 2013, when, in a last-minute decision, President Barack Obama called off U.S. air strikes in Syria,1 has continued to shape his legacy. Instead of striking the Syrian government in retaliation for a nerve gas attack near Damascus, Obama took Russian President Vladimir Putin up on an offer to peacefully dismantle the Syrian chemical weapons program and craft a United Nations resolution2 to make sure no gas attacks ever occurred in Syria again."

https://tcf.org/content/report/red-line-redux-putin-tore-obamas-2013-syria-deal/

At this point, despite all the rhetoric, Trump has yet to be duped by Putin. All you keep pointing out is how Putin utterly fooled Obama. And somehow, in your mind, that makes Trump an idiot.

Comments for this post are closed

Wow, more self-harming hand-waving. This time that Trump is a better opponent to Putin than Obama. Dude, you have to think about how these things will be read by the straights.

Comments for this post are closed

"Dude, you have to think about how these things will be read by the straights."

I suspect the "straights" are going to look at how I made a statement and then backed it up with a specific citation.

Then they are going to look at your response: "Wow, more self-harming hand-waving."

And come to the conclusion that you are a bit of a dumb ass.

I'll spell this out, because it's remotely possible that you don't know what the words you are saying actually mean. Hand-waving is when you object to a specific statement without providing any reasoning. My comment provided not only reasoning, but backed it up with a citation. My response was pretty much the exact opposite of hand-waving.

Comments for this post are closed

By straights I mean normals, people abreast of the news.

You really can't have it both ways. As the Cambridge Analytica thing shows, Trump-Russia love has been a component of Trumpism form the beginning, and as these polls show, it is still bound.

https://www.vox.com/policy-and-politics/2018/1/30/16943786/trump-changed-public-opinion-russia-immigration-trade

It is really absurd for you to think you can prove in a link or two that the world is upside down.

Comments for this post are closed

Maybe you should read Vox more critically (or not at all) because it's rotting your brain.

Did it occur to you that Vox is showing you some cherry picked charts and using that to push a dubious conclusion? To be fair way, below the cherry picked charts they actually show a medium term chart (10 years). And it clearly disproves their thesis.

Here's their thesis:

"Nowhere has this been more evident than in views on Russia. Trump’s presidency has seen Republicans following their leader and warming to Russia and Vladimir Putin, as Democrats report more strongly negative views."

They support it with two charts that start in 2014. Could there be any events that happened in 2014 that might broadly effect American opinions about Russia? Oh yeah. That's right Russia invaded Crimea!

Now let's scroll down and look at the non-cherry picked data. "% who think of Russia as an adversary".

Republicans polled between 20 to 23% from 2009 to 2014. Now in 2017 they poll at ..... 20%. There was a spike corresponding to the Crimea invasion and Ukranian soft invasion. But afterwards it returned to the decade long trend line.

So Republicans have been pretty consistent. How about Democrats? Well they were consistent also, from 2009 to 2016. Roughly 11% to 20%.

When did it drastically change? During the Crimea invasion period it went up from roughly 15-20%. However it's not till 2016 and the election of Trump that it went through the roof. Going from 20% to 38% in less than 2 years. Does it occur to you that the chart mostly shows an over reaction to Russia from Democrats and that the Republicans have instead been pretty consistent?

Comments for this post are closed

JWatts with the kill shot on Obama congratulating Putin in 2012.

But the zombie staggers on: "Back then, Putin was cool!"

Comments for this post are closed

I honestly don't know what anonymous thinks he's doing here aside from humiliating himself

Comments for this post are closed

Comments for this post are closed

https://twitter.com/JoeNBC/status/976261664503824385

Joe's right. Trump should have told Putin to 'cut it out'.

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

The conundrum of 2016: A billionaire who lived in a penthouse decorated like Versailles understood the national mood better than the squads of pollsters, data analysts and professional politicians of both parties.

No matter your political views, this is remarkable. Undermines the assumption that our experts are as smart as they keep telling us they are.

+1, best post in the thread

Yep, +1

Although Trump was just as surprised as the rest of us that he won. A leaked memo at the start of his campaign had him wanting to come in 2nd in the primary and build his brand.

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Yes, they lost a game of strategy to dummy (so they say).

So, who are the dummies?

Comments for this post are closed

+1

Comments for this post are closed

Comments for this post are closed

The left has controlled the MSM (and education and government and NGOs) or, more properly, the MSM is the left! Too bad for them many just aren't eating that dog food anymore. Now Ben Shapiro and Jordan Peterson have massive audiences and the POTUS entertains his fan club with counter signaling tweets! LOL!!!

It is all so delightful to watch! How dreary and dull life would have been had the wicked witch won.

It's true that the Left had pretty much de facto control of the MSM until the 1990's. Talk radio started the trend of allowing an alternate voice to get air time. Then the internet pretty much broke the MSM lock on News completely. The Drudge Report pretty much did this single handedly when he broke the Monica Lewinsky story that Newsweek had decided to quash. The Left will never recover the monopoly they had on the News from 1950 to 1990.

But I think you aren't correct in your belief that they've somehow lost. Too much of the Right leaning news sources are openly biased. So, effectively Fox News squares off against MSNBC and that leaves the middle to CNN, etc. So the Left still controls the middle ground and that will always be the more legitimate position.

Fox News was positioning itself to be in that space, particularly when they shed Glen Beck and gained Juan Williams. But in the past two years they've lost Megyn Kelly and Bill O'Reilly. At this point, the gravity of their programming has swung to Hannity, who has always been a Right wing biased persona.

Have you noticed that all the Left was using FauxNews a few years ago? And now it's stopped. It's because a few years ago, Fox News was a threat to the Left's lock on the MSM and they were attempting to cast it as non-legitimate. Then Fox News made their own mistakes, became less legitimate and now the Left isn't worried about it.

"Too much of the Right leaning news sources are openly biased. So, effectively Fox News squares off against MSNBC and that leaves the middle to CNN, etc."

"Any organization not explicitly right-wing sooner or later becomes left-wing."

"Just over 37 percent of those surveyed said that, given equally qualified candidates for a job, they would support the hiring of a liberal candidate over a conservative candidate. Smaller percentages agreed that a "conservative perspective" would negatively influence their odds of supporting a paper for inclusion in a journal or a proposal for a grant."

Put it all together, add it up and see what is the sum: the left sees their politics as a moral imperative and acts on it in daily life. Only conservatives can operate institutions without political bias and their status is tenuous as there is no countervailing force to leftist bias in an institution with neutral administration.

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

“Unfortunately, when you pursue the feeling of control rather than the actual control, you often end up with neither.”

The biggest and saddest example of that are regulators who, pursuing the feeling of control so much, settled on the risk weighted capital requirements based on ex ante perceived risks, and completely missed out on the fact that what is truly dangerous for the bank system is the unexpected ex post reality.

http://perkurowski.blogspot.com/2016/04/here-are-17-reasons-for-why-i-believe.html

Comments for this post are closed

The issue with facebook is that they could collect data about other people ("you friends") from these apps. Just what kind of data is not clear, and the journalists don't seem to know enough to be specific about this - in theory, they should just be able to get you list of friends, but the analysis makes it sounds as though they got much more... but facebook should not be able to ask you to give away your friend list.

And yet this has been a well published fact since Facebook introduced apps. At this point, not knowing the risks of providing Facebook with personal information and the potential consequences of using Facebook apps requires a level of head in the sand that would amaze an ostrich.

I think you're missing the point. Of course, no one should be surprised FB has a lot of info on him or herself. I think everyone was surprised that someone else using an app could consent to sharing their friends' information (not just the list of their friends - which even that arguably they shouldn't be able to consent to, or else at least the friend should be able to prevent their inclusion in sharing). One person shouldn't be able to consent to the sharing of information of other people (and even worse, (1) I think the consent was "opt out" and (2) it was supposed to be for academic research purposes only).

Facebook says they've stopped that now, but who knows the boundaries of this stuff.

I'm not missing the point. Anyone surprised that apps could gain access to friend lists has simply not been paying attention. This has been a well documented fact since apps came out. There was the usual level of outrage in the media and it was then forgotten. People forget the fact that anything you tell Facebook becomes Facebook's data. You gave it to them and gave FB permission to use it as they see fit. And how they see fit is, again, well documented.

This is similar to what I say about AI and cars below.

Maybe it isn't just that people don't know, it's that they would rather not think about painful details.

Comments for this post are closed

You did it again. "to friends lists" - that's not the point. It's the friends data.

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

https://www.wired.com/story/facebook-cambridge-analytica-response/ has more details - Facebook should never have shared your friends like and should be prosecuted for it. I'm an IT dev, with a strong interest in privacy, and I thought, when I went to use that app that I was selecting for it not to see my list of friends, not their actual likes. I can see why most people didn't pick up on that

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Now let's do it for Global Warming - people are worried about the "Yellow Peril" as represented by the rise of Asia but saying so would be racist so they worry about CO2 instead?

Comments for this post are closed

>Wake up, can't wait to vote for Hilary in a few days
>Check Facebook
>See advert for MAGA hats in purple (my favourite colour, as it says on my fb profile)
>Wtf I love Trump now.

This is what the anti-tech folk really believe.

The problem isn't with facebook, its with giving some types the vote in the first place.

Hard fail.

Comments for this post are closed

Comments for this post are closed

"Could most Americans clearly and correctly articulate exactly ..." anything at all?

"articles about robots putting us all out of work are so popular during a time of full employment": that's "full employment" in a technical sense so narrow that it might as well be a blatant lie.

: "We are about to enter a new meta-narrative for American society, which I call “re-establishing the feeling of control.” "

....psycho-babble from an economist

these Bloomberg columns are mostly low intellectual quality, apparently aimed at making a small media splash with popular audiences via puffed up topics and very subjective speculation (maybe that's where the money is for columnists?)

You left out blatant statism and knee-jerk Trump-ripping, but yes, that's where the money is for economists/opinion-writers (same thing).

Comments for this post are closed

Obviously, the Brexit sloganeering around 'take back control' were just completely coincidental.

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

This is a nonsensical column. We have had exactly one fatality with a driverless vehicle in testing. Look at all the advances in material science and avionics yet we still have plane crashes, albeit infrequently; we still have train crashes as well. As I read the story about the Uber fatality it was a result of something unanticipated (as are some automobile crashes). Maybe further development of software and sensors solves this particular issue the fact remains that there will always be a chance of something bad happening.

With respect to Facebook and Twitter, methinks TC forgets that it's very easy just to opt out and not buy into the alleged Pynchonian paranoia (paraphrasing TC, do read the Proverbs for Paranoids!). I've never had a Facebook or Twitter account but I do worry about how much the analytics of Marginal Revolution are being used!!!!

"This is a nonsensical column. We have had exactly one fatality with a driverless vehicle in testing. "

The gist I took from the column is that we've only one crash, this shouldn't be a big event, yet people have an irrational fear about tech that they feel they don't control (or perhaps don't understand). That doesn't seem nonsensical.

It seems to me that people who know less about AI resolutely ignore those who know more.

https://www.wired.com/2016/03/self-driving-cars-wont-work-change-roads-attitudes/

You have to know who Andrew is and why to trust him.

Comments for this post are closed

Tyler should do a Conversation with Andrew. It would help him and his readers calibrate expectations.

The average person does not know degree to which artificial intelligence and deep learning are oversold by their very names. Practitioners know the great opportunities, but they understand the limitations as well.

This is neat stuff without it really being artificial intelligence or really deep learning.

Comments for this post are closed

One fatality in far fewer miles than the average human driver-miles fatality rate.

We should count the Tesla guy too. He was told he had an autopilot, and hd trusted it too much.

"We should count the Tesla guy too. He was told he had an autopilot, and hd trusted it too much."

If you're going to count the Tesla guy, you'll need to count the Tesla miles, too. More than 1 billion (with a "b") miles have been driven by Tesla cars on Autopilot:

https://electrek.co/2016/11/13/tesla-autopilot-billion-miles-data-self-driving-program/

(Note: This article says 300 million miles driven on Autopilot, but that was November 2016.)

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

"This is a nonsensical column. We have had exactly one fatality with a driverless vehicle in testing."

As Megan McArdle pointed out, given the very high safety of human driving (~1 death per 100 million miles), one fatality within the first 10 million miles, or so, of autonomous car testing is not actually a good record. Especially since autonomous vehicle testing has, so far, avoided the most challenging driving conditions (bad weather, low visibility, snow and ice, rough unpaved roads, etc). Now maybe this death was just chance (like winning the lottery one of the first few times you play), and maybe autonomous vehicles will rack up hundreds of millions of miles before the next fatality. Or maybe not. Would you bet on it?

I am not sure how anyone would grade monitored self-driving, and how many points they would subtract when the human driver takes back control.

It is actually a hard question, because the technology is changing, the software is improving, you hope with every revision human override would be reduced, so how the heck do you set a starting place and start counting safety?

I used to do FDA approved Medical Systems. We completely froze our hardware and software design and tested for six continuous months before release. We did that so we had a hard Baseline of what we were producing and what it would do. The analogous thing for self-driving cars would be to freeze those designs And Then test them for a million miles.

( continuing to love the capitalization of voice input, I guess it is picking up my emphasis )

Comments for this post are closed

I think this comment completely misses the big picture.

Large scale open world testing of autonomous vehicles started in 2016. There's less than 3 years of testing. During that 3 years, the performance indicators have all improved. Even if autonomous vehicles have a 10 deaths per 100 million miles today, that's highly likely to steadily decrease.

Look at the GM Cruise graph in this article. The period it covers is from Dec 2016 to Nov 2017. The improvement is remarkable. (And Google stats are roughly 4x better).

https://arstechnica.com/cars/2018/02/waymo-now-has-a-serious-driverless-car-rival-gms-cruise/

It's quite possible that the technology will plateau, in which case autonomous cars will not be viable. But there's no evidence for that. All the data indicates a rapidly improving product.

There’s less than 3 years of testing. During that 3 years, the performance indicators have all improved. Even if autonomous vehicles have a 10 deaths per 100 million miles today, that’s highly likely to steadily decrease.

Perhaps, but who's going to accept a temporary 10-fold increase in per-mile traffic deaths from autonomous vehicles during some transition period while the bugs are worked out? But also -- tech is full of examples where AI improves at only a steadily diminishing rate and never reaches human levels of performance (for example, OCR, machine-translation, speech-recognition). How much better is Siri than when it was introduced? What's the rate of change? How soon to you expect it to meet or exceed human levels? (I would answer A. 'Not much', B. 'Slow' and C. 'Not in my children's lifetimes'). Notice that what these examples have in common is dealing with messy real-world inputs -- something that definitely applies to autonomous vehicles. This is quite unlike the closed, symbolic 'worlds' of games chess and go where AI has surpassed humans (those are worlds with very simple inputs and rules -- though, of course, the search spaces are huge).

People make an assumption that because autonomous vehicles have 360 degree sensors and aren't distractable, that they'll necessarily be safer. Or that because of steadily increasing processing power, that they'll improve greatly over time. But the problem is not having enough sensor coverage, the problem is making sense of all that data with anything close to human sophistication. And more processing power isn't necessarily going to help (nor is processing power really increasing rapidly any more). Maybe in the Arizona a person would have hit the woman too. But a human driver -- in a glance -- can tell not only is a pedestrian walking a bike, but that -- given the disheveled appearance, odd behavior, and various bags hanging from the bike -- that this is a homeless person who may be under the influence or have mental health problems or both and that they should be watched very closely and given a wide berth. Being able to make that kind of judgement instantly and automatically (and avoiding accidents even when they wouldn't have been at fault) may be a big reason why human drivers are as safe as they are. And autonomous vehicles cannot do that. They can't do anything at all like that. They have a hard time telling what objects ARE let alone telling what they MEAN and what they are likely to DO.

You what I want? Before letting an autonomous system take the wheel, I want my car to be equipped with a passive 'vehicle narration system'. I want the AI to be able to give me a stream of running commentary from the 'passenger seat':

"I'm currently tracking 25 pedestrians, Dave. 19 of them are out of range of our vehicle path and can be ignored. Of the 6 remaining, three are adults who have made eye contact and have clearly seen us. One is leaning against a post and talking on the phone. That one seems unlikely to move, but they're oblivious so let's watch them closely. And two are kids who are playing a game and aren't looking at us at all, so we should slow down."

Does that seem like science fiction from the distant future? Well, it *shouldn't*. If an autonomous driving system can't interpret the scene to that level of sophistication, it has no business piloting a two-ton vehicle through an environment filled with pedestrians.

That there are driving situations where human drivers outperform AI drivers does not imply they are better overall. There may be many more situations where not being distracted, drunk, tired or having a fast response time is very advantageous. This can then result in self driving cars still being a lot safer than human driving ones. Insisting on a perfect record would not be rational. And btw people on drugs and playing children also get hit by human drivers sometimes. Humans also misread ambiguous cues . You see this at 4 way stops all the time.

Comments for this post are closed

"Perhaps, but who’s going to accept a temporary 10-fold increase in per-mile traffic deaths from autonomous vehicles during some transition period while the bugs are worked out? "

The states of California and Arizona obviously.

Comments for this post are closed

"The states of California and Arizona obviously."

Provisionally. But if, as autonomous vehicle operations become more common, accidents start piling up at higher rates than for human drivers, that permissive approach may end abruptly.

Comments for this post are closed

Comments for this post are closed

Even if autonomous vehicles have a 10 deaths per 100 million miles today, that’s highly likely to steadily decrease.

If they are 10 times as dangerous as human drivers right now (and based on the Uber fatality it looks more like 20x), they are more-or-less equivalent to drunk or sleep-deprived drivers. So get them off the road.

Comments for this post are closed

Comments for this post are closed

" given the very high safety of human driving (~1 death per 100 million miles)"

Human driving is extraordinarily dangerous. That's one death per .0000000052 parsecs of driving.

Here's another stat: about 1.3 million people die every year in car accidents. That's not "very high safety." That's horrific. Yes, you made a good point responding to the post, but let's not forget that transportation safety is a gigantic problem.

Compared to even relatively recent human history (going back no more than a century and a half), the orders of magnitude of improvement in both speed and safety of travel is a miracle. When you're talking about a global population of 7 1/2 billion, you can get large numbers even with very low rates (as in this case). Not to say that more progress isn't possible or desirable (it is -- especially in developing countries), but treating traffic safety as a 'horrific', 'gigantic' problem is over-the-top.

Comments for this post are closed

Comments for this post are closed

"Now maybe this death was just chance (like winning the lottery one of the first few times you play), and maybe autonomous vehicles will rack up hundreds of millions of miles before the next fatality. Or maybe not. Would you bet on it?"

I'll be happy to bet that autonomous vehicles will reach 100 billion vehicle miles traveled before the first 1000 fatalities. (Meaning that the rate over the first 100 billion miles will be less than 1 fatality per 100 million miles traveled.)

Say...$50?

I'd also be willing to bet $50 that "autonomous vehicles will reach 100 billion vehicle miles traveled before the first 1000 fatalities".

I'd take both bets. And I'd be happy to lose (but I don't think I would). We would need some reasonable deadline (5 years? 10 years?), and if autonomous vehicles fail to rack up that many miles in the time-frame -- because of safety problems or regulatory limits imposed because of safety problems or manufacturers realize they're not ready for primetime, then I'd win.

The more we know about the Arizona crash, the worse it looks. The woman was walking slowly across an open area in front of the vehicle. It was dark, but that does not matter for LIDAR, and the car did not slow or swerve or give any indication of tracking her at all. It's pretty clear, too, that the 'safety driver' wasn't attentive to the road or surroundings and wasn't adding anything to the margin of safety.

Comments for this post are closed

"I’d take both bets. And I’d be happy to lose (but I don’t think I would). We would need some reasonable deadline (5 years? 10 years?), and if autonomous vehicles fail to rack up that many miles in the time-frame — because of safety problems or regulatory limits imposed because of safety problems or manufacturers realize they’re not ready for primetime, then I’d win."

I couldn't agree to the time limit, unless it was something like 2030 or later. Congress and regulators can introduce all manner of hurdles completely unjustified by science. (For example, marijuana is *still* classified as a drug without any currently accepted medical use.) The only thing that would overcome it would be China having most of its fleet being autonomous, and Congress and regulators would be afraid of being completely left behind.

But if you're willing to go with a deadline of 2030 or later, I'll do it. I'll post the bet on my blog:

http://markbahner.typepad.com/

"The more we know about the Arizona crash, the worse it looks. The woman was walking slowly across an open area in front of the vehicle. It was dark, but that does not matter for LIDAR, and the car did not slow or swerve or give any indication of tracking her at all. It’s pretty clear, too, that the ‘safety driver’ wasn’t attentive to the road or surroundings and wasn’t adding anything to the margin of safety."

Yes, I absolutely agree with that, after seeing the video of the accident. It's especially bad that the woman who was hit was crossing from the median, so she'd already crossed at least one lane. Plus, with the road being so wide, there were plenty of places to swerve and still stay on the road. Finally, the road was deserted, so slamming on the brakes would not have caused an accident.

But this accident, and the reaction to it, actually demonstrate how *safe* autonomous vehicles will be. I'm sure that just in the last year, of the 6000+ pedestrians killed by automobiles, there were hundreds of accidents like this. But the *human* fleet learned nothing, so that this year the exact same types of accidents will kill just as many people. In contrast, the *autonomous* fleet will be improved dramatically from this very terrible accident. Not only will Uber/Volvo improve their entire fleet, but other autonomous vehicle manufacturers (GM, Waymo, Ford, BMW, Tesla, etc.) will likely improve their systems...just from this single awful fatality.

Comments for this post are closed

I can only believe after reading these comments that JWatts and Mark Bahner are not software developers. Because the absolute glibness of these replies indicates that you have no idea what complex systems programming is like.

Comments for this post are closed

"I can only believe after reading these comments that JWatts and Mark Bahner are not software developers. Because the absolute glibness of these replies indicates that you have no idea what complex systems programming is like."

That's right, I'm not a software developer. But I do know a fair amount about:

1) Historical/future trends in computer hardware (e.g., processors, memory).
2) Developments in computer vision hardware (e.g., lidar, GPS).
3) What autonomous vehicle manufacturers are predicting for trends.

I also understand why autonomous vehicles are likely to lead to transportation-as-a-service, and how that feeds back to enhance the development speed and safety of autonomous vehicles. And I understand that fleet learning with respect to safety is orders of magnitude more rapid with autonomous vehicles than it is with human drivers.

If you think I don't know what I'm talking about, I'd be happy to also bet you $50 that autonomous vehicles travel 100 billion miles in the U.S. before the first 1000 deaths (i.e. a death rate of less than 1 person per 100 million passenger miles).

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

When I was a young man about 20, another young man came up to me and asked me to sign a nominating petition to get a candidate on the ballot. I asked who? He said a name I don't remember, but the person was running to be the communist party candidate for some office. I signed it.

I had no desire to see the person elected. I had no romantic notions of communism. My feeling was that if allowed to speak freely, allowed to enter the community of ideas; society would reject communist ideology. For me, at that time, they were just another voice wanting to be heard. Indeed the nuns who had taught me at a much younger age often sounded like they had communist leanings, except the nuns rejected the atheism stuff. (Liberation theology had an attraction for some of the nuns.)

Perhaps I was naive to think that our society is healthy enough to listen to competing ideas and to choose wisely. Perhaps I had too much faith that it is difficult to fool or manipulate the masses in the long run. I preferred to think that our institutions were strong enough to withstand any passing fad ideology and that a free society that allowed dissent, even stupid ideas, was the best way to remain free.

When I hear people say that we must stop Russian interference in our elections, I wonder how that will work. How do we prevent Russians from manipulating our institutions that encourage free debate without limiting our ability to enter free debates? Our social platforms are easily abusded. Domestic organizations routinely try to use them to shape, even manipulate, us. Politicians continually give us poll tested speeches that can conflict with what they want to do or what they believe.

But today I do worry about the deep state and their partnership with a corrupt media. I worry that privacy and our ability to express our ideas openly, without fear of retaliation, is declining. We have always had periods where skilled manipulators could gain some power at the head of a mob. But this time it feels different. The control of the state has increased. We have grown more dependant on the power of the collective. What was once a marketplace of ideas has become, it appears, a shadow society which seeks greater power through the manipulation of our fears. Perhaps 1984 just arrived a little later than Orwell predicted. I hope I'm wrong.

BTW for those interested in Russian manipulation of elections here is an article that claims the KGB helped invent and spread Liberation Theology in Latin America.

https://www.nationalreview.com/2015/04/secret-roots-liberation-theology/

You aren't going to get any bites on giving 1 single shit about Russian involvement in environmentalism, liberation theory, or communism/socialism from the left. This matters because:

"How do we prevent Russians from manipulating our institutions that encourage free debate without limiting our ability to enter free debates?"

Limiting dissemination of conservative thought on social media is the goal.

To clarify my point, which was unclear. It has been alleged by former communist intelligence operatives that the KGB was very involved in Liberation Theology way before the internet was invented.

Their ability and desire to manipulate organizations and disrupt governments has been occurring globally for decades. It appears to be difficult to prevent in free societies. The internet may give them more bang for the buck but isn’t really a new policy for Russia.

Comments for this post are closed

Thomas
Your comments are very powerful and compelling. This should be screamed from the rooftops.

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

While I sympathize with Cowen's emphasis on loss of control, there is a very big difference between loss of control when one flies on Delta and loss of control when one gives social media one's personal data. I recall learning to fly (in a Cessna 150) 50 years ago, and how I felt much less vulnerable when I was flying the aircraft than when my instructor was. Of course, that was irrational. As for autonomous cars, I'm amazed at the level of enthusiasm for them. Indeed, whenever I remind people that Google's engineers said they would have to be limited to 30 mph to be safe as long as they have to share the road with non-autonomous cars, the reaction from readers of this blog is that I'm just a Neanderthal and don't understand tech. On the other hand, if autonomous cars are provided a separate right of way, then they wouldn't be limited to 30 mph. But wouldn't autonomous cars traveling on a separate right of way just be "transit" by another name? Today, Ross Douthat has decided that it's cable and broadcast news that is to blame for Americans' loss of control over the election of Donald Trump. Russians, social media, cable and broadcast news, blame anybody but the Americans who voted for Trump. Douthat might have blamed the large cohort with whom Douthat identifies: evangelical Christians. But that wouldn't be fair since evangelical Christians aren't in control, God is.

" I remind people that Google’s engineers said they would have to be limited to 30 mph to be safe"

Drink!

Comments for this post are closed

"whenever I remind people that Google’s engineers said they would have to be limited to 30 mph to be safe as long as they have to share the road with non-autonomous cars, the reaction from readers of this blog is that I’m just a Neanderthal and don’t understand tech. "

In case anybody was wondering, rayward has been asked for a citation on this comment repeatedly and I don't believe he's ever been able to back it up. Also, his numbers have been inconsistent over time. It's clear that Google/Waymo runs their prototypes at fairly low speeds. However, I've never seen a Google engineer state that autonomous cars will be limited to 35 mph permanently.

Reference:

rayward 8/21/2016 "That means one of two things: “self driving cars” that have maximum speeds of about 25 mph (what Google envisions) or “self driving transit” that is an alternative and requires a separate right of way."

rayward 9/8/2016 - " I’m fine with the hype, provided it is accompanied with a little reality, in particular that self driving cars will be limited to about 25 mph (according to Google’s engineers). "

rayward 01/03/2018 - " Google’s engineers have stated that autonomous cars have to be limited to 30-35 mph if they share the road with non-autonomous vehicles. "

Comments for this post are closed

The quote by the Google engineer was in a NYT business article several years ago. A Google search will find it. I would add that I've read maximum speeds ranging from 25 mph to 35 mph, which explains the different speeds in my comments. Whether 25 or 25 isn't important, since even at 35 mph what's the demand for such a slow car. But logic and common sense would cause any rational person to realize that autonomous cars will have limited speeds if they share the road with non-autonomous cars. I'm all for autonomous cars, but for them to be practical, someone will have to build a separate right of way. And you know who that will be. Why not just improve transit since autonomous cars traveling on a separate right of way is "transit".

"The quote by the Google engineer was in a NYT business article several years ago. A Google search will find it."

And yet I wasn't able to. If you can't provide a citation, then most of the readers are going to continue to doubt your statements.

I did find this quote. It does not back up your claims.

"“We’ve capped the speed of our prototype vehicles at 25 m.p.h. for safety reasons. We want them to feel friendly and approachable, rather than zooming scarily through neighborhood streets. Like this officer, people sometimes flag us down when they want to know more about our project"

https://www.nytimes.com/2015/11/14/business/google-driverless-car-is-stopped-by-california-police-for-going-too-slowly.html

Comments for this post are closed

And misremembering one Google engineer 'several years ago' doesn't make your quotes above any more valid. Obviously you are simply wrong about this hobbyhorse, and that's ok. No one is right about everything, except of course Art Deco.

Comments for this post are closed

"But logic and common sense would cause any rational person to realize that autonomous cars will have limited speeds if they share the road with non-autonomous cars."

I guess I don't have "logic and common sense" then. People are already traveling more than 1 million miles a day when the computer is driving the car at highway speeds. They might disengage every couple hundred or thousand miles, but the autonomous car is sharing the road with human drivers at highway speeds.

+1 billion Tesla Auto-pilot miles and climbing rapidly....

Comments for this post are closed

Comments for this post are closed

And yet, working near an area where all the autonomous cars frequent, I regularly (i.e. at least once or twice a week) see Waymo cars driving on the freeway at 65 mph. They also frequently drive the local streets at 45 mph. So apparently Alphabet's employees don't currently believe their cars need to be limited to lower speeds, as empirically, they don't limit them.

Be very, very careful crossing the road then.

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

The authorities have issued a preliminary judgment that the accident in Tempe was the result of an errant pedestrian, and that it had nothing to do with the car in question being driverless. Still, if you think that Americans process information about risk rationally,

Processing information about risk rationally doesn't necessarily involve blindly trusting the authorities; the authorities may not have skin in the game.

Comments for this post are closed

“ Unfortunately, when you pursue the feeling rather than the actual control, you often end up with neither”

I would love to hear some suggestions on how we CAN get control back because I believe you are admitting here that we have indeed lost some.

Comments for this post are closed

It is a bit comical it only took four words in the above essay ("an erratic American president") to ignite so many erratic comments.

You become what you choose, boys.

Erratic in a nutshell:

https://www.axios.com/white-house-leak-trump-putin-call-congratulations-53c68b0d-7e3f-4211-9816-fd3c24e927dd.html

Comments for this post are closed

Goodness.

https://twitter.com/ReformedBroker/status/976433496595075072

Comments for this post are closed

Comments for this post are closed

Re: your 12/2017 Sunday assorted link re: decline of guitar sales (for which comments are closed) NB

https://reverb.com/news/guitaronomics-have-guitars-become-more-expensive-over-time

Comments for this post are closed

Just a couple years ago there was campaign which slogan was literally "taking back control".

I would say it's not a new meta-narrative. It always have been there, sometimes dormant, sometimes wins a referendum.

Comments for this post are closed

"The first American citizen to be assassinated by a mini-drone on U.S. soil will be a big, big story."

I'm selling anti-assassination hats for the low, low price of $299.99, available in a wide assortment of colors!

That's a lousy deal, these anti-assassination hats are only $15.

https://www.teeshirtpalace.com/products/i-m-with-her-hillary-2016-trucker-hat?gclid=EAIaIQobChMIuePz_Mz92QIVVVcNCh0QiQWbEAQYASABEgJcgPD_BwE

I’ll take two of each! Especially since I’m really worried about autonomous drones, not the guided ones.

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

"It resembles how a lot of people feel in greater danger when flying than driving a car, even though flying is usually safer. "

I don't think fear of flying is about control -- I think it's about being tens of thousands of feet in the air traveling at very high speeds. People on buses, trains, and passenger ferries have no control either, but I don't think that causes them fear as flying does.

My grandfather owned a plane and hated flying in large airliners because of the lack of control.

He sold his plane later because he thought it was too risky.

Well, he was right. His plane *was* risky. Private aviation is much less safe than driving (which, in turn, is much less safe than flying commercial). As for lack of control in an airliner -- he was a special case there (relatively few people are pilots).

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed

Prof. Cowen, with all due respect, since when did we ever have control over data or technology or anything in that matter? I believe it's an illusion. (I am not saying we shouldn't but I am saying it is almost impossible.) BTW, I published an article (Jan/2013 Analytics Magazine) which examined how big data, analytics, social & digital technologies had changed 2012 elections and political campaigns similarly to how data-driven approaches had changed the sports of baseball. http://analytics-magazine.org/big-data-analytics-and-elections/

Yes, we are "fooled by randomness".

Comments for this post are closed

Comments for this post are closed

Its not that complicated. The media wants to sell ad and they sensationalize everything to attract attention. If they only engage in readers or audiences who think critically, they would turn off everything else who makes up the majority.

Comments for this post are closed

Like drones, driverless cars possess some features of an especially potent scare story.

1) Well there is a comfort of knowing you are 'in control' in general. Look at fear of flying which is not new and bizarrely safe.
2) I think driverless cars are still 10 years away. They have a good driving record in the right situations but still have a long way to go. Their impact will be limited if they are not good driving in the rain and poor conditions. I don't think they get fully integrated until 20 years. (We often forget how long technology breakthroughs take to become everyday products. Remember the internet was invented/discovered in 1969.)
3) Again, what CA did was unauthorized by the users data and tended to a lot of Pizzagate crap. (CA is probably legally in the clear from the government but might be liable to Facebook. However, any lawsuits by Facebook may open them up to a lot of publicity.)

4) Probably the main point here is how much are we controlled by outside forces. Equifax royally gets hacked and we have to spend $60 to block our credit. (And better yet how have the business leaders of Equifax suffered?) If we click on certain items on Facebook then we get endless steams of political material.

Comments for this post are closed

"Most of all, driverless cars by definition involve humans not feeling in direct control. It resembles how a lot of people feel in greater danger when flying than driving a car, even though flying is usually safer."

People are not in control when they are passengers in a car and yet aren't anxious like many plane riders.

Comments for this post are closed

The dual of 'giving up control' is 'trusting some other entity with control.' But why should we trust Facebook/Google/Amazon/Uber/.... ? Because the free market will eventually fix things? Because regulators will stay ahead of the curve and keep these companies in line before anything bad happens?

Comments for this post are closed

People have long had lingering fears about the profiles that data companies have on them. (And with good reason.)

There's nothing unusual about the CA story, except that people don't like Trump (and with good reason), and maybe this is someone they can blame, since the other blame games don't seem to be working, and Zuck really does seem like a jerk, so this is a bunch of smaller storms finally merging together.

I doubt anything will change. People will post angrily about it, on Facebook even, and then move on with their lives. It's a shame and I wish it were different, but it won't be. The things that would actually make a difference would cost something, like losing the convenience of Facebook.

Comments for this post are closed

Just the facts. Human drivers kill one person per about 11 million miles.

The statistics for automated cars (all test miles) is in the same ballpark. However, each computer mortality will probably be the last from that source, if it was a preventable accident. The same is not true of human drivers who repeat the same errors like drunk driving again and again.

Note that the human emergency driver didn't prevent this accident. Human backup to computers doesn't work.

Note the deadliest drone on this planet is the mosquito and look at the size of its CPU. It just tracks CO2, heat, and smell to find you.

"Just the facts. Human drivers kill one person per about 11 million miles."

Human drivers are involved in accidents that result in a fatality at the rate of approximately 1.1 fatality per 100 million miles.

Comments for this post are closed

Comments for this post are closed

Comments for this post are closed