December 29, 2018

"Now do : large female doctor pushing a stroller with a cat in it."



Link to tweet.

71 comments:

gilbar said...

it's not supposed to be Preferences! It's Supposed to be POINTS!
Death Race 2000

Tommy Duncan said...

Another headline: "Organizers cancel Women's March Jan. 19 due to 'overwhelmingly white' participants"

"EUREKA, Calif. — The organizers of the annual Women's March have decided not to hold a rally in Eureka on Jan. 19, as previously planned, because they say participants do not represent the diversity of the area."


Perhaps we should poll regarding a large white female doctor?

Temujin said...

Nice system. Keep working on it. BTW- who decides what is 'old'? The computer or some 23 year old? Hey, how about putting a category in for 'activist'?

Ann Althouse said...

"BTW- who decides what is 'old'?"

We're going to get face lifts so cars won't run us down.

Ann Althouse said...

I noticed that women were preferred to men in every category except old.

Ann Althouse said...

No. I'm wrong. The male doctor was also preferred to the female doctor.

rehajm said...

Conspicuously absent is ‘vehicle occupants’. That one’s the deal breaker for most peole and will make the chart irrelevant.

rehajm said...

Tuck it in so the car can’t see, men.

Ann Althouse said...

Here's your new handbag. Carry it wherever you go, Doctor. The car won't know you've only got your sandwich and your iPad inside.

traditionalguy said...

Rating our Driver's inner hates . It must be from a new Scientific Dating Site. Either that or a part of the new Driver's License Test.

Ralph L said...

Squirrel isn't on the list because it would ruin the scale.

You'd think the white coats would obscure the sex.

I used to think self-driving cars were a crock and a sales gimmick, but many young people don't seem to want or like to drive, if my niece is any indication.

Kevin said...

So it doesn’t have time to save everybody but it has time to calculate age and occupation? Based on what? Do all doctors wear lab coats on the streets? Executives wear what, top hats, mourning coats, and monocles?

Or is everyone going to be microchipped so the car can scan your chip really fast? Will doctor chips flood the black market?

I want to see the comparison between doctors and nurses. That ought to bring some quality REEEEEEEE-ing.

Clyde said...

Why do foreigners hate cats? (Americans don’t spell it “favour “.)

Ralph L said...

Size matters.

Virgil Hilts said...

As many have noted, real danger with all the AI/robots is not so much they'll go rogue or be poorly programmed as that they can/will be hacked and/or reprogrammed. ISIS attacks on Christmas markets with self-driving lorries. SciFi in 2019/evening news in 2024.

CJ said...

4chan already solved the hierarchy of persons problem with the famous “Switch Track” philosophy.


https://m.imgur.com/OgdZDN4

Dust Bunny Queen said...

In my self driving car.....the self being MYself, I try not to hit anyone or anything.

However, if it comes down to hitting a cat, dog or any other type of critter, I'm not going to swerve and crash my car. Grit my teeth and feel bad about it. A deer, cow or even a bear...yes, I will swerve as the mass of those animals versus car will cause some major injuries to myself and my vehicle and I might be able to avoid the hazard if I act fast (This is a common occurrence BTW)

My onboard computer....my actual brain, can make that decision in a split moment.

However, I have no idea how you can tell a Doctor from a regular person who is in front of your moving vehicle.

Fritz said...


Blogger Tommy Duncan said...
Another headline: "Organizers cancel Women's March Jan. 19 due to 'overwhelmingly white' participants"

"EUREKA, Calif. — The organizers of the annual Women's March have decided not to hold a rally in Eureka on Jan. 19, as previously planned, because they say participants do not represent the diversity of the area."

Perhaps we should poll regarding a large white female doctor?


Diversity in Eureka? That's a good one...

tcrosse said...

The preferences take in age, gender, and size, but not race. How nice.

JAORE said...

I suspect wearing a MAGA hat will result in an intentional hit and run.

AI strikes again.

Lucien said...

How do you determine the racial makeup of a march that hasn’t happened? Was there a signup in which people were required to state their race?

Fernandinande said...

People who think other people's worth depends on their physical characteristics should be the first to die.

Dust Bunny Queen said...

Fritz....good catch. :-)

Have they ever BEEN to Eureka?

The area is whiter than Wonder Bread. With the exception of a rather large Portuguese continent from the dairy farming family industry south of Eureka.

SayAahh said...

“Doctor “
As in MD or PhD?

Henry said...

What about pointy-eared dog vs. floppy-eared dog?

Henry said...

If self-driving cars can identify criminals, the TSA should use them in airports.

Fernandinande said...

I noticed that women were preferred to men in every category except old.

One hundred and thirty countries with at least 100 respondents were selected (range, 101–448,125). ...

Distributions across the three clusters reveal stark differences. ...

For example, countries belonging to the Southern cluster [C. and S. America] show a strong preference for sparing females compared to countries in other clusters.

Fernandinande said...

If self-driving cars can identify criminals,

Some of the groups - apparently the most interesting ones - were put in just for fun. Literally:

"Additional characters were included in some scenarios (for example, criminals, pregnant women or doctors), who were not linked to any of these nine factors. These characters mostly served to make scenarios less repetitive for the users."

FIDO said...

Hey, how about putting a category in for 'activist'?

We are designing crash avoidance, not a targeting system

Henry said...

As I understand the AI question, the real answer is "brake for all of them." An AI that is trying to distinguish between multiple categories of input -- Bear? Doctor? Shadow? -- is sucking milliseconds away from a decision. If something is in the road, you don't try to figure out what it is. You brake.

The problem is that no one will be happy going somewhere in a self-driving car that brakes for shadows.

* * *

I had a short conversation with my son today about safe driving. I was going between 35 and 40 on a 35 mph road. Ahead of us I saw a woman on the sidewalk with a dog on a leash. They were standing and looking away from the road at a pond. As I approached the woman turned as if to make a step up the sidewalk and I immediately slowed down. My son asked me why I slowed down. I told him that once the woman turned to move, I was less sure of what she might do and I was really less sure of what the dog might do, so I wanted to be going a little slower until I passed them.

This is something an AI can figure out without much problem. But the more things like this it is figuring out, the slower its real-time reaction and the higher the chance of a false positive or false negative.

JHapp said...

I will going to a garage sale soon to get a stroller or two.

FIDO said...

I would expect that the Feminists would cede their AI privilege and want the programmers to remove any female avoidance preference...in keeping with their so called principles.

But you never see women agitating for 50% representation in garbage collection. Odd.

wild chicken said...

"You'd think the white coats would obscure the sex."

Doctors don't even dress like doctors anymore except in surgery. How would you know?

The only doctor I ever saw in a white coat was the pot doc. What a joke. "You're gonna LOVE cannabis! Heheheh."

chuck said...

Quiet, I'm depreciating.

Fernandinande said...

The "Women's March" has been having trouble with their own racists and racism for a while, this cancellation might be a reaction to that.

Wiki: "The racial makeup of Eureka was 79.3% White," which is not too different from the US as a [?]hole. (76%?)

But Eureka is in short supply of the most important race: 1.9% African American[sic].

? = "w" or "shit"?

Leland said...

Who is giving the self-driving car moral agency? Can we kill that person first?

exhelodrvr1 said...

Henry,
"But the more things like this it is figuring out, the slower its real-time reaction"

Relatively speaking, it's insignificant - milliseconds at the most, and the typical 2-3 second reaction time that humans have essentially disappears.

Roy Lofquist said...

An Irishman, a doctor, an albino lesbian and a penguin cross the road...

Henry said...
This comment has been removed by the author.
Hannio said...

"An Irishman, a doctor, an albino lesbian and a penguin cross the road..."

I remember that album cover.

Matt Sablan said...

I want to know how it identifies. Or by criminal do they mean jaywalker like in the article?

Henry said...

@exhelodrvr1

The AI loses time if it is waiting for more input because it doesn't have enough data to make a decision. The latency is in the input, not the processing.

Dust Bunny Queen said...

Another issue about relying on an AI to drive your car or relying on your own experience is.....you have experience.

For instance....this area of the road is where the deer usually cross (suicidally leap out in front of your car) between the hours of X and X. So you are more aware because of this past experience.

Plus you know from past experience when you have missed or waited for the first deer, that there will be several more racing out to hit you in the SIDE of your vehicle. You stop and wait.

The kids in this neighborhood street are notorious for playing IN the road and often dash across to get their ball. Sometimes on purpose.

The field along this stretch of highway has a lot of migratory birds which will fly in front of your car because the hawks are hunting. You can SEE the birds hundreds of yards in the distance and get prepared.

I know...mostly rural examples. BUT the fact that you know these things from historical experience and an AI would not know.

Experience trumps.

Matt Sablan said...

Like, look at the widget at the criminal line. That's a guy carrying a purse, not a jaywalker.

Matt Sablan said...

Dust Bunny: If anyone built an AI for roads and didn't use actual data, then it was poorly built AI. A properly built AI will not only tell you when deer are likely to cross, but will also tell you every reported deer crossing on record; when crashes occurred with deer that are reported, etc., etc.

AI will always aggregate more experience than any one person, provided the AI is built/coded properly. Now, there can be glitches in the code, sure, but just like it is incredibly rare for a person to outsmart a machine at a game of pure skill once the machine has had enough time to learn the game, the AI will, eventually, have more experience than a driver, provided good data is being fed to the AI.

Matt Sablan said...

"Conspicuously absent is ‘vehicle occupants’. "

Included in the data set total, just not that screenshot.

Dust Bunny Queen said...

every reported deer crossing on record

LOL Sweet Summer Child.

Self driven (AI) cars have their place. They can probably be a good thing in some areas. Cities, major highways. Through Musk's boring tunnel. Have at it.

On a road that most GPS programs can't find, or which send you off into a field, isn't it.



Matt Sablan said...

Actually, cities are probably where you'll get the least amount of use out of them with stop and go and random pedestrians. Over a long road, you'll have:

* All data from tagged deer (which there are a lot of)
* All reported accidents involving deer (and non-deer)
* All reported incidents of a deer on the side of a highway
* Self-reporting by users/drivers

All of that will trump any single driver's experience.

And that's before you actively start monitoring the road and the area. Will we get there in any time soon? Probably not in the next year or so -- but in time, the AI is going to trump us.

Dust Bunny Queen said...

* All data from tagged deer (which there are a lot of)
Where? I've never heard of such a thing where I live. And I've hunted deer. I do believe that the antelope might be tagged.


* All reported accidents involving deer (and non-deer)
* All reported incidents of a deer on the side of a highway
* Self-reporting by users/drivers


LOL again.

Seriously? You think people OFFICIALLY report that they've seen a deer on the side of the road? You think that we report that a deer jumped out in front of our car and we missed the deer? Live deer..... we just are thankful that they didn't leap out in front.

Who are you going to report TO anyway. Hey County official..I saw a deer!!!! Wow. Fish and Game....there are deer in the woods!!!! The deer are all over my property...can you guys do something about that??? They would think you are crazy.

In cold weather it is an issue because they like the salt that is used to keep the ice off of the road, so usually the deer just stare at you as you drive slooooowwwwly by and continue eating. No one reports this...WHY would we. To whom? The County or Cal Trans for chumming for deer?

Dead deer? We take our tractor or a tow chain hooked up to a truck and drag the dead deer off of the roadside to keep from hitting the vultures or coyotes eating the carcass. No one is going to come and clean it up if we told anyone. We might report the dead deer to Fish and Game, but they don't care as long as you aren't going to "salvage" the road kill for yourself Then they care. If it is on a major highway....Caltrans or the County might care of it and THAT might be reported. The County doesn't care either.

Yeah. If your car is totaled or damaged by a deer, you are going to report it to your insurance company. If the damage is slight, less than your deductible, you just shrug and go on with life.

For computers, remember garbage in/garbage out. Incomplete data=stupid AI

Seriously. Self driving cars do have their place and that place isn't everywhere.

Matt Sablan said...

Not like a live, grazing deer. But dead ones hit in hit-and-runs.

SGT Ted said...

I love it when progressives inadvertently reveal their eugenic and morally relativistic preferences.

Dust Bunny Queen said...

Not like a live, grazing deer. But dead ones hit in hit-and-runs.

Then you are going to be missing a lot of pertinent data.

Matt Sablan said...

That's why I said it isn't something to happen any time soon. But, just like computers became way better chess players than the average person, there is going to be a time they're way better drivers too.

Yancey Ward said...

Where did Sarah Connor fit on the list?

RobinGoodfellow said...

Blogger Dust Bunny Queen said...
Fritz....good catch. :-)

Have they ever BEEN to Eureka?

The area is whiter than Wonder Bread. With the exception of a rather large Portuguese continent from the dairy farming family industry south of Eureka.


Since when are Europeans not white?

gilbar said...

DBQ (i think,) said... I'm not going to swerve and crash my car. Grit my teeth and feel bad about it. A deer, cow or even a bear...yes

The latest conventional wisdom is: Don't Veer for Deer
'They' 'say' that you're MUCH more likely to get into an accident trying to avoid a deer than by just breaking and continuing straight at it.
This seems crazy, until you consider:

The Deers are actually pretty good at jumping out of the way
The Deers are actually pretty good at jumping RIGHT into your new path
The thing you REALLY do NOT want is an out of control skid into the ditch
The thing you REALLY REALLY do NOT want is an out of control skid into oncoming traffic

Of Course, if you break hard, you've got to worry about the person behind you.
(The last deer i hit, i was braking (in car), but Very worried about cycle tailgating me)
(but, the deer before that, that hit MY cycle (and gave me my Traumatic Brain Injury (damn deer!)), i'm pretty sure i was veering (since, that's what i always did,) i don't know though, since i have no memory of the two weeks around the accident)

gilbar said...

RobinGoodfellow said...Since when are Europeans not white?

Portuguese aren't pumpernickel; but they sure aren't Wonderbread. More of a whole wheat, i'd say

Dust Bunny Queen said...
This comment has been removed by the author.
Dust Bunny Queen said...

Since when are Europeans not white?

Since they decided that it is racist to discriminate against Spaniards from Mexico, Muslims from Serbia, Italians from Brazil, the Roma (gypsies).

When Rachel Dolezal can be black, anyone can be anything.

Define white.

ALP said...

Kevin: Executives wear what, top hats, mourning coats, and monocles?

********
!!!! Coffee-spewing funny! Executives look just like the Monopoly Guy.

ALP said...

So if the car is presented withe a crowd of people that includes half of these types, does it find the one lowest on the hierarchy and hit that one? Or does it see the baby in the carriage and treat the whole crowd as such?

ALP said...

DBQ:

Self driven (AI) cars have their place. They can probably be a good thing in some areas. Cities, major highways. Through Musk's boring tunnel. Have at it.

On a road that most GPS programs can't find, or which send you off into a field, isn't it.
************
Your post is inspiring a "Black Mirror" fantasy in my head in which a self driving car gets confused in a rural area, locks the occupants inside, and mills aimlessly in a field somewhere until the occupants persih.

William said...

The right kind of algorithm will do much to help us solve the problem of obesity in America.

Rusty said...

I'm with Gilbar on this.
Is there a league for this? Is Las Vegas taking bets?

jimbino said...

Looks like terrorists will start to carry their explosives in a stroller instead of backpack, vest or pressure cooker.

Hey Skipper said...

[DBQ:] On a road that most GPS programs can't find, or which send you off into a field, isn't it.

Exactly.

AI works great, until it doesn't work at all. And it can't tell one from the other.

exhelodrvr1 said...

Henry,
I guarantee the AI will not wait for input. It will be designed to act based on how the input it has is interpreted by the rules that have been coded.

I would guess that self-driving cars would eliminate probably in the upper-90%s of accidents. Think about what causes most accidents - driving too fast, getting distracted, reaction time too slow. All of those will be virtually eliminated as causes. There will be a VERY small range of conditions that some human drivers would handle better than the computer.

However, I still am not a fan of the concept.

Openidname said...

Ladies, when the self-driving cars arrive, start tucking a pillow under your shirt.

Henry said...

@exhelodrvr1

We don't disagree. The devil is in what's been coded. I'm basically summing up some of the critiques I've read of self-driving cars.

Assume the AI is processing visual data (bear, shadow, baby) in which the worst outcome (not shadow) is evaluated as highly improbable at any given moment, but also recognized as high risk -- and so the AI slows down for shadows. If the AI is coded to wait for more input based on certain probabilities in play, it will make a more accurate decision, but at a later point in time.

In the future, people who have never driven for themselves may not care. But in the present, human drivers accept high risk. Human drivers consistently drive faster than is safe and consistently ignore risk -- slippery when wet; deer at twilight; limited vision. The safest AI won't drive through a misty crepuscular forest faster than 10 mph.

Bunkypotatohead said...

I guess "lawyer" would be off the scale, below cat.

Char Char Binks, Esq. said...

Google recently had a tough time distinguishing Blacks from gorillas. I sympathize.