nathan_long

> If I had a dollar for every DUI story I've heard that involved getting pulled over "a block from home" after rolling through a stop sign..

If a cop suspects DUI, I want that driver pulled over even if they're entering their own driveway. "But I wasn't going to put any more people's lives in mortal danger by my recklessness tonight" is not a good argument. And keeping people from recklessly endangering others, whether by penalties that make them think twice or by taking their license, is one of the most basic reasons we have police at all.

> What the author calls "remarkable abilities" and "robustly understand[ing] the world" can also be seen as just reproducing our own innate and learned collective human biases.

From the article:

> “The bareheaded man needed a hat” is transcribed by my phone’s speech-recognition program as “The bear headed man needed a hat.”

If A.I. is going to work with humans, then yes, reproducing some of our "biases", such as the tendency to describe someone as "bearheaded" far more often that "bear headed", is a good goal. In an alternative universe of chimeras, a different bias would be needed.

But in any case, the A.I. needs to "understand" that one of these transcriptions is much more likely to be correct than the other. So a level of processing beyond "sounds like this word" is needed.

Take the famous absurd quote++: "Time flies like an arrow. Fruit flies like a banana." Reading the second sentence makes you, a human, double-take, and re-read the first sentence, and try reinterpreting it to see if it makes sense a different way. (The fact that it doesn't is what makes it funny.) An A.I. that has no "wait did that make sense?" step, no "is that funny?" step, is going to produce results that disappoint or confuse us humans who do.

(++Interestingly, this quote seems to originate in A.I. research: https://quoteinvestigator.com/2010/05/04/time-flies-arrow/)

True. People can always find a reason to justify their belief or disbelief if they really want to.

I agree in principle, and making an ID easier to get makes sense.

However, in my state (Georgia), they're being strict about photo ID and name matching, yet have ignored for years serious security problems with the electronic voting machines which leave no paper trail.

This makes it easy to believe that the ID requirement is about something other than securing the election from tampering.

How does any of that prevent someone from looking at the voting records, seeing that Bob never votes, filling out an absentee ballot in Bob's name, and dropping it in the nearest mail pickup box?

"1 in 4 Statisticians Say They Were Asked to Commit Scientific Fraud"

It would be hilarious if this report were debunked on the grounds that they inflated the number of statisticians surveyed

Proving the legitimacy of a photo or video is a great problem to try to solve in the age of "fake news" accusations and extremely powerful editing software.

> It's your responsibility to make sure you trust that entity, and do your due diligence.

What does this entail? I mean, ask the average person if they trust the New York Times, the London Stock Exchange, or Spotify. Those are well-known names - sure we trust them. We trust that, as an organization, they are not plotting to steal our identities.

But trusting them means trusting their business people, their IT people, and their advertising partners, not only to be moral but also to be competent. And whoops, all of them have served malvertising in the past.

Nobody has the time and expertise to evaluate every site's JavaScript every time they visit. The "due diligence" you describe would be a very specialized full-time job.

Whereas turning off JavaScript in the browser takes about 10 seconds.

> Damn right you don't let random people drive your car! just like I expect you not to click every link you see!

The difference is that I can interact with any number of people, yet only trust a few to drive my car.

You seem to suggest that I should trust everyone I interact with to drive my car, and since that's risky, I should limit my interactions to a few people.

> You're shifting the responsibility from yourself to a different entity for the choices you're making.

Not trusting a site's code is taking responsibility for what it might do to me. Trusting their code is giving them the responsibility to do what's right.

> The single safest thing you can do while using the web is to simply be aware of what you're clicking on, and what sites you visit.

This is good, practical advice for non-tech people. But fundamentally there is no reason the web needs to be unsafe. Viewing a series of hypertext documents is safe, no matter who wrote them. The web is totally safe if you browse it using curl. It's only when we assume that "of course you're going to run that stranger's code" that things become dangerous.

It's great that we can have web applications. It's not great if every page is an application. There should be an explicit step from "I'm casually looking at your page" to "now I want to trust you and run your code".

> That said, modern browsers do a really, really good job at isolating the code running in that page from anything you care about.

That code can't wipe my hard drive, but it can track me everywhere if I let it.

The reason we need content security policies and cross-site request forgery tokens and cross-site scripting protection etc, etc is that our browsers are constantly eval'ing whatever code they're given.

> That said, I'd be willing to wager a fair bit that literally every line of code you've run on your machine (probably ever if it's been bought in the last few years) outside of the vendor installed OS and drivers came from the internet.

We explicitly decide to install software and we know where we're getting it from. We may not be careful enough, but I certainly trust `brew install` a lot more than I trust a random site I happen to visit.

> You bought that machine to run code. If you don't want to run code that sites serve you on the internet, don't visit them.

I bought my car to drive it, but that doesn't mean that every person I pass on the street gets to drive my car.

You seem to think that visiting a web site establishes a trusting relationship with that site. I think that downloading a site's HTML and running its JavaScript are quite different levels of trust. Especially when the content and the code come from very different entities - eg, the journalist writing the article vs the advertising network the publication uses.

Clearly if I'm using Gmail, I'm trusting Gmail and I don't mind running their JavaScript. But the blanket statement that "if you don't want to run code that sites serve you on the internet, don't visit them" seems out of touch with how pervasively nasty the internet is.

Taking a walk on a city sidewalk does not require eating whatever you find there.

You're not wrong, but...

Distinguishing "quality" content from "engaging" content does not sound like a simple problem at all. It sounds like one that involves a whole lot of heuristics and opinion.

If it were easy, "clickbait" wouldn't be a thing.

I would say it is our own human inability to separate "engaging" from "quality" that causes this mess.

Includes an explanation of exclusion constraints. These let you enforce things like "no overlapping reservations of the same property" without race conditions, similar to a unique constraint.

> Seems like a perfect thing for "market forces" to decide

I'm not sure about that. Consumers can easily tell which phone is less expensive, which one has more storage, etc. It is much harder to know, much less decide based on, which ones will last a long time. Especially since by the time we can confidently say "these last 5 years", you can't buy that model anymore.

This is a problem with nearly anything more complex than a kitchen knife.

This isn't about distribution, but someone once wrote about getting PostgreSQL upsert performance to be better than MongoDB's by disabling some of PostgreSQL's safety features.

https://markandruth.co.uk/2016/01/08/how-we-tweaked-postgres...

This made me laugh. Snarky interpretation: "yeah, we can go fast and sloppy, too. We just usually don't."

> It is a crime under French law to intentionally shorten the life of any product in order to promote sales. The French consumer protection agency has the power to fine up to 5% of annual turnover or impose a jail term.

Nice law.

> This interpretation hinges on interpreting successful sub-majority writes as not necessarily successful: rather, a successful response is merely a suggestion that the write has probably occurred, or might later occur, or perhaps will occur, be visible to some clients, then un-occur, or perhaps nothing will happen whatsoever.

> We note that this remains MongoDB’s default level of write safety.

This sounds pretty scary. How does it compare to other distributed dbs, like Riak? My understanding is that Riak lets you specify how many nodes a write must succeed on to be considered successful. Are its responses more reliable? Is this just a "distributed computing is hard" situation?

Every time I see ads in a new place, I think of this quote.

> Leela: Didn't you have ads in the 21st century?"

> Fry: Well sure, but not in our dreams. Only on TV and radio, and in magazines, and movies, and at ball games... and on buses and milk cartons and t-shirts, and bananas and written on the sky. But not in dreams, no siree.

How efficient is laser weeding in time and energy?

Like, can the robot move down the row as fast as a human weeding by hand? Can it carry a car battery and weed for an eight hour shift?

The pitch from Small Robot Company's home page (https://www.smallrobotcompany.com/):

> We’re an agri-tech start up commercialising a deceptively simple idea: small robots not big tractors.

> Because unfortunately big tractors are neither efficient nor environmentally friendly. Currently, 95% of energy is used ploughing. And ploughing is only necessary because of heavy machinery crushing soil.

> We are building robots that will seed and care for each individual plant in your crop. They will only feed and spray the plants that need it, giving them the perfect levels nutrients and support, with no waste.

> This level of detail allows you to be kinder to soil, kinder to the environment, more efficient, more precise and more productive. It’s the best of all worlds. An increased yield, as well as minimal chemical usage. So you can increase revenues by up to 40%, reduce costs by up to 60%.

> Our robots are being designed and built by farmers, for farmers. Because we have spent the last year talking to you, we know you don’t want to buy robots. You are worried about the cost, and what happens if they break down. Instead you can lease our robots through a Farming as a Service (FaaS) model.

Obviously it may not be as good as it sounds. But it sounds amazing.

Also these quotes from the article about scale were interesting:

> The agility of agricultural robots means small farms with compact fields will no longer be at a disadvantage; independent shops and restaurants will be able to grow their produce on smallholdings efficiently tended by Rachel-like machines.

> ...

> “Most people think this is going to be expensive, is going to do everyone out of a job and is going to be good for the big farms, not the small farms,” he says. “It’s actually the exact opposite. The big farms are all about economies of scale: big fields, big tractors. We are developing small machines. I believe the extra production we need to feed the planet is going to come from small farms that can’t use those economies of scale.”

I love the idea of a farmer on 2 acres competing with Big Ag on price.

I don't understand your comment. Suppose the robot is the size of a person or smaller. Does a person damage the ground by walking across a field?

I never said that. It's not perfect, but I far prefer living in a capitalist society to a centrally-planned economy, which I think can never work.

The game of capitalism requires referees. Otherwise, robbery is an excellent business model.

> Is there anyone that enjoys sprinkling every bit of SQL with upper() comparisons?

Does using a citext column not solve this?

One "quality of life" thing - you can wrap changes to the schema in a transaction, so that a given set of changes is atomic. (Elixir's Ecto and Ruby's ActiveRecord both have migrations that do this by default.) As far as I know, that doesn't work in MySQL.

There's also neat stuff like exclusion constraints: http://nathanmlong.com/2016/01/protect-your-data-with-postgr...

And there's the ability to do all kinds of neat geospatial queries with PostGIS - eg "list all properties within X distance from the curvy coastline".

> In a capitalist society (such as we live in, like it or not), we express value in money. IF your company isn't making money, is it adding value?

Yes, but it's a very imperfect way to capture value.

Without legal support, it doesn't account for negative externalities - eg, if your factory saves money by dumping waste into a stream, making people sick (but you don't pay for that). Or your site gets visitors by sending spam, wasting the time of the recipients and everyone who works trying to filter spam (but you don't pay for that).

It doesn't account for things consumers don't have the foresight to care about, or don't have options to choose between. Eg companies that manufacture a fridge that will last 10 years and be unrepairable may make more profit than those who would make one that lasts 50 years and be repairable, because they'll sell more product. Planned obsolescence wastes resources, but is profitable.

It also doesn't account for the value of things like raising children. If I play for my own children for 2 hours, that's not part of GDP, but if I pay someone else to, it is.

And it doesn't account for the fact that people's desires and priorities are warped. Child pimps generate revenue, whereas the social workers rescuing kids from them do not.

> One thing I really like about Twitter is that you can choose to follow who you want. If you don't want to see memes making fun of you, just don't follow accounts that post/retweet these memes.

My experience of Twitter is the opposite. Someone I follow posts something I think is intelligent and polite, and the replies are filled with people launching tangential attacks and making weird accusations.

A tweet is like a tiny conference talk whose attendees are self-selecting. Yet random strangers all over the world are able to detect the in-progress talk, parachute in, and yell crazy things, and nobody can take away their microphone.

> It's sad that Twitter is banning satire accounts. I'm against all censorship of lawful speech, and condemn Twitter and any social media platform that participates in this kind of censorship.

Would you distinguish satire from impersonation? Eg, https://twitter.com/RealDonalDrumpf?ref_src=twsrc%5Egoogle%7... is clearly self-identifying as a satire of Donald Trump. Nobody is going to confuse that account with the real one.

But suppose hoards of people with political opinions you detest start creating accounts that pose as "your side", using the worst possible arguments for your positions, insulting everyone, and making "your side" look stupid and cruel. Is that a kind of speech you want to protect?

My first iPod Nano with a scroll wheel. I could adjust the volume and scrub back/forward through a podcast without looking at the device. All it did was play music and podcasts, and it did it very well.

Now I have an iPod Touch. It's heavier than I want in my pocket for running. It has a bunch of apps besides music which I don't need, and which necessitate it having a lock screen. And ironically, because it has a "touch screen", you can't use it solely by touch.

Not https://en.wikipedia.org/wiki/Palm,_Inc., maker of the Palm Pilot. :(

> your real taste (and what they can get you to binge) vs. the taste you want people to perceive you as having via ranking

You make this sound like a question of honesty vs pretense, but "what they can get you to binge" reveals a different set of motives. It's sometimes a question of self-control.

I want to be healthy. I also want to eat junk food. What do I "really" want?

Imagine an online grocery store which noticed they could profit more by hiding healthy options from me and selling me more junk.

Would their AI be acting according to my "real taste"? In a sense, yes. Would it be serving my best interests? No.

If Netflix can entice someone to binge watch TV all weekend, does that mean that's what they "really" wanted?

Every AI serves someone's interests. It's important to ask whose.

And if you want to make an AI that truly serves your users (rather than exploiting them), give them explicit control. Stop with this "we know you better than you know yourself" nonsense. Let them say things like, "yes, I watched that movie / bought that candy, but I wish I hadn't. Please don't use that for future recommendations."

Good point. And the fact that it's posted by a VC might have something to do with the content of the advice.

There can be death spiral here, though.

  Them: "We're shutting down X because too few people use it. Also, announcing Y!"
  Users: "I'm not using Y. You'll probably shut it down."
  [months pass] 
  Them: "We're shutting down Y because too few people use it."

The brand recognition that they hope will make Y successful might actually work against it.