GPS Navigators, morality and the human utility function

April 2, 2014

GPS navigators used to be simple. You plug in the desired address, they calculate the quickest route. At some point, many started getting a feed of traffic information, so that they can route against big jams. Still pretty simple.

Enter Waze. Waze hyper-micro-optimizes for traffic conditions, using other users as roaming sensors and, probably, complicated AI on the back-end to predict what route would take how long. This means Waze learns from experience how fast people drive where, under what conditions — and, as far as I can tell, takes all this information into account. This means that on a residential street that is constantly being speeded to, Waze will learn that it is a fast path, and will start directing drivers through it. Drivers who are using Waze, which means they want to get to their destination as fast as possible.

There are a lot of valid reasons to want to get somewhere fast. Maybe you scheduled a meeting. Maybe someone is having an emergency. Maybe you just want to cuddle your kid before she goes to bed. Waze does not care. It tries to get users to their destination as fast as possible, no matter which residential streets they speed through. It has no concerns for safety, either the driver’s or pedestrian, neatly divulging itself of all responsibility by “just suggesting”. What if a nice street starts being higher-up on the Waze paths? And this lowers property values? Waze has no concerns for that.

I’m not picking on Waze especially. Their only fault was that they took the optimization criteria that all GPS navigators use (find fastest path) and used superior sensors and superior algorithms to implement them better. Whoever wins the “navigator wars” would have had to do at least as well, so in that sense, there is nothing special about Waze except hiring smart people who made good decisions.

But this does show how AI moves forward: smart engineers optimize the hell of whatever target function they are given. At some point, it starts having real costs that humans would understand, but an AI would not care about because the AI cares about optimizing the target function. The only way to solve it is to figure out what humans care about in terms computers can understand and make sure that any AI takes those into account — even if it is not smart enough to self-modify, it can do plenty of damage without it.

In other words, the Friendliness apocalypse, if we include not just existential risk but also a general raising of human risk factors, is not in some nebulous future until we figure out self-modification. We are in the middle of it, and we should make sure that we take it into account when building AI — even if that AI is limited to “just suggesting things”, because humans are suggestible, and have learned to trust computers’ suggestions.

Public letter to Congresswoman Jackie Speier, CA14

April 16, 2013

Dear Congresswoman Speier,

I have already called your office, so I am sorry if this is redundant, but they did not have an answer for me. I am sending you this message because of my concern about HR 624, the “Cyber Intelligence Sharing and Protection Act”, also known as CISPA. It appears that it would be up for a vote in Congress soon. I am worried for my privacy, as well as for the privacy of other Americans, were this bill to be passed. This law violates due process by allowing private companies to hand data to the “government” even when they are under contract to not do so, without any court order. I am sure most people working for Federal agencies are good and honest people, but this allows one bad apple to grab whatever data they need for whatever private means they have with no judicial oversight. It also allows private companies to hack into computers as long as they believe “it is necessary” — in essence privatizing law enforcement, again without proper judicial oversight.

In the wake of the tragedy in Boston, the US must, of course, be ever-vigilant against threats. However, laws like CISPA, making our country more surveiled and less free, are essentially give terrorists what they want — to terrorize us into hurting out essential freedoms.

Moshe Zadka

Babies — a guide

April 13, 2013

Some people have asked me to summarize the collection of links I posted about baby development. Here is my unauthorized summary. There are links to the original resources throughout the post.

A lot of the following information is based on, or inspired by, the Educaring philosophy as put down by Magda Gerber and taught by RIE. Educaring philosphy is based on trust in the child to initiate development, giving children safe areas to explore in an uninterrupted way, involving the child in all child-care activities, observing the child and responding to her needs and being consistent when limits are concerned. The central theme is respect for the child as a whole person in his own right.

When babies are born, they have one mode of communication — crying. Therefore, a lot of what parents do in the first few months is dealing with crying babies. The blogosphere, unsurprisingly, reacted with a wealth of information to the question “what do I do with a crying baby?”. The basic newborn care advice — check diaper, burp, and feed — only goes so far. If your baby is not giving you hunger signs — clenched fingers and fists over chest and tummy, flexed arms and legs, rooting, fast breathing or sucking noises or motions — feeding might stop the crying even if it is not the best idea.

After those, comes the “unexplained crying”. “Unexplained”, of course, is not a fact about the crying — it is a fact about the parent. Babies have feelings too and need to express them.

The diagnosis of “colic” as long stretches of unexplained crying is giving way to the concept of “PURPLE” crying. PURPLE stands for “Peak” (of crying) “Unexpected” “Resists soothing” “Pain-like face” “Long lasting” (several hours a day) “Evening” (also known as “the witching hour”, around twilight). Apparently all mammals go through their own equivalent of PURPLE crying, the human animal between 2 weeks and 3-5months.

Listening to the crying, being compassionate without trying to stop it, is important. Crying is a form of communication. When we do want to soothe the baby, CALMS1 is a popular approach. [C]heck in with yourself, and connect with your feelings — you want to calm yourself before calming the baby down. [A]llow yourself to take a breath and relax. [L]isten to what your baby is trying to tell you. [M]irror — empathize with the baby’s feeling, voicing his emotions as best you can. [S]oothe your baby. It is OK if the soothing initially makes the crying harder — the baby might need to voice the hurt feelings first. Sometimes, the end result of this approach is to “hold the baby through the tears”, being calm and letting the baby cry as long as she needs to. When soothing a baby, pick a tactic (singing, bouncing, etc.) and stick to it — switching tactics is a bad idea. Repetition, repetition, repetition. There are reasons to be wary of the “5S”.

Parents need to understand that sometimes the baby will cry, and this does not mean they are failing as parents. Unfortunately, that is exactly what attachment parenting is teaching.

Second only to crying in the amount of energy parents spend obsessing about is the baby’s sleep. Nursing kids to sleep does not make them sleep longer — putting them to sleep earlier does. Althought not too much earlier, and not too soon. Wait for them to display sleep signs, and put them to bed then. Babies’ “long-stretches” of sleep start from 3 hours at newborn, through 4 hours at two months, 4-5 at 4 months, 6 hours at 6 months and by 1 year old, 80% of babies sleep through the night. Routines, especially at 4 months and beyond, are useful for helping babies sleep. Babies are woken up by hunger, dreams or discomfort. Some babies fight sleep, and need to be helped there. SIDS is mostly a risk while sleeping, follow AAP recommendations to reduce risk: put babies on their backs to sleep, use a firm sleep surface, room-sharing but no bed sharing and keep soft objects and loose bedding out of the crib.

A distant third is how children play. Even babies can do independent play as long as they are placed, carefully, on their back until they have learned to roll to their tummies. It is OK if the baby does not “do” anything — just kicking their legs and staring is fine. The area, of course, must be safe — a no-”no” area, where everything possible is permissible. It is OK, and even good, to stay near the child and observe — but even if they seem frustrated, not to help them (this is like giving someone the solution to a riddle). Babies are perfectly capable of entertaining themselves this way. Babies can start having play-dates early on — if we believe them capable, if we are attentive to their moods, make it into a routine, and are patient with them. As always, observing them in quiet is important. Tummy time is important developmentally. Start slow, get down to your baby’s level and provide entertainment while the baby is on his tummy. Avoid having tummy time right after feeding, or when the baby is upset. Keep it calm and quiet, so the baby is not scared by what she cannot see. Be patient — the baby will eventually want tummy time.

Parenting Resources

March 4, 2013

Like any normal (read: insane) new parent, I have been reading a lot about babies online. Below is a reading list of sorts. I do not agree with everything here. Some of the things I do agree with, I still do not follow. It is not unbiased (there is a definite NVC/compassionate parenting bias). It is not complete — if you have more resources, please link in the comments!

Compassionate parenting posts


Play and activities




Parenting styles

February 7, 2013

So far, I see three competing “parenting philosophies” that are popular around these parts.

First is the style I will call “mainstream”. As examples of this style, I will take Dr. Karp and Dr. Mindell. That style is based on getting the children to behave appropriately within the family. Karp’s specialty is calming down babies, Mindell’s is getting them to sleep well. This style approaches babies somewhat “behavioristically” — it revolves around sending in the right inputs to achieve the desired results.

Then there is attachment parenting, as shown in the Sears baby books. This style revolves around an 18 year long pregnancy, in essence. Babywearing, home-schooling and nursing the baby whenever they want to be nursed.

The last style is NVC parenting, based on Magda Gerber’s teaching (and also shown in “How to Talk so Kids Will Listen and Listen so Kids Will Talk”). This style revolves around the metaphor of a relationship with the baby — essentially, welcoming them as a member of the family with their own responsibilities and needs.

It is easy to confuse attachment parenting and NVC parenting. The easiest way to tell them apart is to see how they treat crying. Attachment parenting sees itself as a solution to crying babie. In NVC parenting, crying is accepted as a valid form of communication, and there is no attempt to stop the crying — but to understand the crying, validate it and respond to underlying needs if possible.

Because of obvious ethical concerns, it is hard to do randomized intervention to determine which method leads to good results. However, based on studies like  “Decomposing the Genetic Variable in Income: The Role of Cognitive and Noncognitive Skill,” MIT Working Paper, 2010 and “What Do Twin Studies Reveal About the Economic Returns to Education?” The American Economic Review, Vol 85, No. 3, it seems like parenting has little effect on lifetime income (which is at least a reasonable proxy measure for “life success”). This means parenting should not optimize for life success, but for happy family life throughout childhood. So that would mean, do the behaviorist thing at the beginning, and then less and less in favor of the NVC style as the preferences of the child become more and more salient.

Essential skills for the 21st century

January 8, 2013

In the late 19th/early 20th century, the western world had a literacy revolution. Everybody, or most everybody, could read and write. Now, most people did not become scholars, but they could read an article. Most people did not become novelists, but they could write a note that they’ve gone to the grocery store. The amount of value we, as western society, got from this universal literacy is unimaginably big — even though most people do not make their living off of reading and writing. Similarly, thanks to Leonardo de Pizza (AKA Fibonacci), we all can do arithmetic — we can add 154 to 289. Although we have computers to do many calculations now, it is still useful to be able to do simple calculations (how much am I paying for an ounce in two different detergent brands?) Again, the amount of value we got here is unimaginably big — even though now nobody makes their living from arithmetic for a living anymore.

I think in the 21st century, we need to add a few more skills to that roster. First of all, programming. Programming is the literacy of the 21st century. In a world filled with computers, knowing how to automate tasks, and having an accurate model of how software works, makes people better off. I think every kid should learn at least one high-level programming language, to the point where “Send a form letter to all the e-mails mentioned in a text document” should be seen as the equivalent of writing a quick note. While I do not expect everyone to be able to come up with novel solutions to Prisoner’s Dilemma-type situations, understanding the basics around the math of decision making — something like the stuff mentioned in “Smart Choices” — would get them to make better decisions, be it about choosing a major in college or deciding whether to buy a house. Though I cannot point to a concrete benefit, I think teaching the basics of Calculus and Linear Algebra would do a world of good. At least to the point of being able to cover the basics of Quantum Mechanics so that they understand, on a deep gut level, that the world is made of math. That discovery, originally mentioned by Galileo (in The Assayer), shocked the world once upon a time — what does math have to do with nature – and I sometimes get the feeling people still do not, in their hearts of hearts, really believe that.

2012 — a retrospective

January 5, 2013


This was quite a year.

In February, I got married. That was a pretty big deal.

In March, I gave a well-received talk in the biggest PyCon ever.

In April, I went to Israel for a 2nd wedding reception, and to Italy for my honeymoon.

In May, I discovered I’m an expectant father.

In October, I moved to a house — back-yard, front-yard, the lot.

In November, I got my green card.

Somewhere along the way, five of our friends had their first kid, which is kind of amazing. Babies!


Get every new post delivered to your Inbox.

Join 290 other followers