ctl

Software-only machines are qualitatively different than machines with paper trails, since they can be tampered with en masse by a small number of adversaries without needing physical access at voting time; and because such tampering may not leave any trace whatsoever.

Do you really not see how that’s a big deal?

It's politics. He's saying it because it's an effective thing to say, not because he actually didn't know health care is complex.

So your philosophy is, "I only do healthy things if they're also painful"? That doesn't seem productive.

You’re right. I didn’t read the paper very carefully, and was myopically focused on the emulating-a-real-brain AI strategy. As in, let’s slice up a real human brain, map the neurons and synapses, and then simulate them as faithfully as possible.

To do that you need a great deal of fidelity in your simulations of neurons, which are enormously complex. But there is an argument to be made that neuronal complexity is incidental to the brain’s overall “computational capacity”; that you could replace the neurons in a human brain with much simpler nodes and still end up with a functional intelligence, after sufficient rewiring.

I don’t think that claim is obvious, but it’s definitely possible. And if it’s true, you can have human-level intelligence for 10^19 ops, given suitable software.

So I apologize for my post. It was over the top and unfair.

All that said, I still disagree with Bostrom’s conclusions. I think he enormously understates the difficulty of creating intelligent software, if we’re not just copying an existing brain.

Let's look at the most important section of the paper. He estimates the processing power of the brain:

The human brain contains about 10^11 neurons. Each neuron has about 5 • 10^3 synapses, and signals are transmitted along these synapses at an average frequency of about 10^2 Hz. Each signal contains, say, 5 bits. This equals 10^17 ops. The true value cannot be much higher than this, but it might be much lower.

In other words, there are 5 * 10^14 synapses in the brain, and each synapse transmits up to 100 signals per second, and we can probably encode each signal with 5 bits. That's ~10^17 bits per second.

So, uh... does anybody else notice that that's not an estimate of processing power?

That's an estimate of the rate of information flow between neurons, across the whole brain.

The level of confused thinking here is off the charts. Does this guy not understand that in order to simulate the brain, you not only have to keep track of information flows between neurons, you also need to simulate the neurons themselves?

That's not merely a flaw in his argument. It indicates that he has no idea what he's talking about, at all.

Needless to say, this paper and its conclusions are complete nonsense.

Well, if it's possible to build a human level intelligence, it's probably possible to build an intelligence that's much like a very smart human except it runs 100x faster. And in that case, somebody with sufficient resources could build an ensemble of 1000 superfast intelligences.

That's a lower bound on the scariness of AI explosion, and it may already be enough to take over the world. Certainly it should be enough to take over the Internet circa 2015...

To my mind it seems pretty clear that if AI exists, then scary AI is not far off.

That said, I don't worry about this stuff too much, because I see AI as being much technically harder, and much less likely to materialize in our lifetimes, than articles like this suppose.

My understanding is there are two reasons you want inflation:

(1) Inflation means real interest rates can go negative, which means it's harder to get stuck at the zero lower bound (as we are today).

(2) Wages are sticky, i.e. workers fight very hard against having their wages reduced. When there's inflation you can effectively cut wages by giving raises below inflation. Without inflation it's very hard to cut wages, and the economy becomes brittle.

That's exactly why it's called a deflationary spiral: once you're in it, it's hard to get out even with aggressive monetary policy.

My general impression is that more and stricter regulations in an industry is usually (though not always) a bad thing for consumers in that industry

In cable internet the average consumer has only one or two potential providers. On top of that, companies who try to break into the industry face significant political and regulatory challenges when they lay fiber.

I don't understand how you can think this kind of industry should be unregulated. There are few market forces keeping providers in line, and we see the results of that in the industry's attempts to extort internet companies with throttling.

This is astonishingly high quality for pop science. It deserves more upvotes!

You're complaining about the system... in response to a powerful instance of the system working correctly? What?

No, booking a ride with the intention to cancel is definitely wrong. My point is that Uber employees aren't doing that frequently enough to affect Lyft at all, so this can't be a plan on Uber's part.

It sounds like some Uber employees are just being assholes for no reason. That's unfortunate, but it's no reason to malign Uber itself. Every company has some assholes.

Lyft says that Uber employees have canceled 5000 rides since last October. That's like $100k in lost revenue -- it's nothing. Definitely not some big plan by Uber to undermine Lyft.

Uh... which of Uber's actions here are unethical?

Aggressively poaching drivers from Lyft seems all right to me. That's just healthy competition.

And this stuff about wasting Lyft's money by booking and canceling rides -- it's a bunch of nonsense. By Lyft's own estimate, Uber employees booked and canceled 5000 rides since last October. That's like $100k/yr in lost revenue. It's nothing. Certainly not a systematic attempt to undermine Lyft.

Am I missing something here?

Disclaimer: of the two services I only use Lyft. I've had bad experiences with Uber's customer service.

Your comment makes no sense. GDP is slightly down compared to 2013. Why are you talking about 2008?

The article implies that the main driver here is reduced consumer spending from a harsh winter.

If the fixed income came from newly-printed money then yes, it would cause inflation. If it came from taxes then no, it would not cause inflation in and of itself.

In both cases, it would increase the purchasing power of the poor and reduce the purchasing power of the rich, because $5000 means more to somebody making $10000 than it does to somebody making $1000000.

Other industries are deeply unlike tech, so comparing salary trends would be meaningless.

A sibling comment suggests that we look at salaries in tech companies that didn't collude. But that doesn't work either, because (1) those salaries could have been indirectly affected by the collusion, and (2) other tech companies are different than Apple, Google, et al -- most importantly, they have less money!

If, in my nonfiction book, I write, "Under normal circumstances, water boils at 80 degrees celsius," then I am a liar, and there is nothing I can say after the fact that will excuse my behavior.

Malcolm Gladwell often writes lies in his nonfiction books. I have found myself personally deceived by some of his lies. For instance, on page 39 of Outliers, he writes:

The striking thing about Ericsson’s study is that he and his colleagues couldn’t find any “naturals”, musicians who floated effortlessly to the top while practicing a fraction of the time their peers did. Nor could they find any “grinds”, people who worked harder than everyone else, yet just didn’t have what it takes to break the top ranks.

The first time I read this passage, I thought it was one of the most extraordinary pieces of information I'd ever encountered. It meant that practice almost completely determines ability! I mostly believed that for more than 4 years. (And so did the rest of the world -- that's the core of the 10000 hours meme.)

It turns out that the passage is completely false. Ericsson's study did not actually say anything about individual violinists at all -- it only ever reported average statistics of groups of violinists.

Other studies of the relationship between total practice time and ability have found that different people may reach a given level of ability with enormously different practice totals. For instance a study of chess players found that the average amount of practice time needed to reach the Master level is ~11000 hours -- but the standard deviation of the distribution is more than 5000 hours. One person only took 3000 hours to reach Master level. Another took 23000 hours. [1]

Calling this kind of deception "being playful with ideas" is absurd.

[1] Note: I have not read the study in question. My info comes secondhand, from http://www.sportsscientists.com/2011/08/talent-training-and-... and http://www.outsideonline.com/outdoor-adventure/media/books/H....

Right now people avoid self-employment because if they get seriously sick it will bankrupt them. Under the new law it won't.

That's the argument. The premiums you pay when you're healthy are beside the point.

If nobody's willing to do the job for $6K then why would the salary of the job drop to $6K? Nothing you've written makes any sense.

Citation? I really like the idea of basic income but $2700/month per person seems absurdly high. I don't know why any political faction would try for so much.

1) you might as well not give any money to anybody, because relatively no ones wealth will change

Blatantly untrue -- under basic income poor people will see their relative purchasing power increase, and rich people will see theirs decrease.

3) you will end up with inflation, because free money will create more demand, which in turn will result in smaller supply (relatively to demand) and higher prices

Or maybe supply will go up? You know we have an unemployment problem, right? More demand is exactly what America wants right now. It creates business and employment opportunities.

in the end these $1,000 after a couple of months will buy you only $500 worth of goods (end game is that these $1,000 will be worthless after some period)

Obviously basic income would be pegged to GDP, hence pegged to inflation. This is a non-argument.

This is an incoherent argument.

The people who successfully make money have a larger portion of it taken away by the government and given to people who don't make money. They recoup their losses by raising prices and re-extracting it from the people it has been given to

That doesn't happen in the real world -- raising taxes doesn't cause runaway inflation. (How could you possibly think that it does?)

BI has to be increased to keep pace with rising prices caused by funding BI [...] The endgame of the second option is Marxist communism

Even if raising taxes did cause runaway inflation, this would still be wrong. Say that each year you set aside 20% of the GDP for basic income. Then your basic income plan already automatically increases with inflation.

So the percentage of GDP used for basic income would be stable. We would never get communism.

The endgame isn't the Soviet Union, it's Denmark.

Citation?

I love this article, but: what can a practicing math teacher take away from it? How can you apply this stuff if you still have to teach a standard curriculum?

I'm really asking -- my friend is about to start as a high-school math teacher.

I guess the first recommendation would be: motivate every new technique by starting with one or more problems that the technique helps to solve. (Here "problems" is meant in the Lockhart sense -- real puzzles, not exercises.)

But how often are "techniques" actually taught in high school math, especially algebra and precalculus? A lot of high school math consists of digesting new definitions, or the generalization of old definitions. A fair amount of it consists of learning theorems that go unproven, or that are proven (by the teacher) too quickly for students to understand where they come from -- and in general it isn't satisfying to solve a puzzle with a theorem that one doesn't actually understand.

On top of that... students have to spend time with problems before they become genuinely interested in their solutions, so progress would be slower with this method. It's not clear that you could teach a whole year's curriculum in one year like this. (And if you fail to do that you'll eventually get fired.)

Any insight? I believe that it's possible to teach math, even standard high school curriculum, in such a way that students are at all times intrinsically interested in what's presented. But it would be awfully hard to do at scale, at the standard pace, as a high school teacher would have to. How might a teacher start in that direction?

But then these people look at history, and see this giant pile of anecdotes, and all the sudden it's the Middle Ages again and we can hermenutic our way to wisdom by staring at it really hard

My whole life is a pile of anecdotes! And yet I've used my experiences to build a functional mental model of the world. I can use history in the same way.

You argue that anecdotal data should be flatly ignored when making decisions. If that were true then newborn children would be the best reasoners on the planet. Your argument cannot hold in general.

So basically: overpopulation is bad, therefore human death is a good thing, therefore curing disease is bad.

This argument is simultaneously (1) viscerally horrifying and (2) completely wrong.

It is wrong because curing disease does not increase population growth rate. Birth rate is extremely well-correlated with infant mortality rate, and pretty well anticorrelated with both life expectancy and quality of life. That's why the countries with the highest population growth rates also have some of the highest mortality rates. If you want to stop overpopulation, you should target common causes of death, especially those that afflict young children.

Which is exactly what Bill Gates is doing.

After I wrote this one, I had to go back and armor it by pre-empting anything I could imagine anyone willfully misunderstanding to use as a weapon in comment threads. The whole of footnote 1 is such armor for example.

You don't give your detractors enough credit. Many of these misunderstandings are not willful, and actively seeking to avoid them is almost always good practice on your part. (In my opinion.) For the record, I found footnote 1 illuminating.

I'm sure that the intense scrutiny you get would annoy any writer. But I think it genuinely makes your essays stronger, too, and I hope you don't hate it too much.

Let me tell you: if you were the one doing the design of a motion sensing system, I can bet it wouldn't be "easy".

This sentence makes no sense. The patent in question is not on a single technical implementation of pinch-to-zoom; it is on pinch-to-zoom in general. It does not merely protect the (easy) engineering needed to detect two fingers pinching, it protects the idea of "two fingers pinching = zoom". This is an obvious idea, and should not be protected by patents.

I think it's an overstatement to say that online programs are the future of education. Coursera and Udacity are not selling the same thing that universities are selling, at all, and I find it hard to imagine them affecting higher education substantially.

What do people gain from spending four years in college? Many of them learn something, but most of the value is in the degree, which serves as a signal of competence in the workforce and is a requirement for most jobs. Online programs probably won't provide the same kind of signaling in the foreseeable future. Moreover even students who are primarily interested in learning are unlikely to get as much out of online programs as they do out of college; the rewards structure of college drives many students to work much harder than they would on their own, and (without the signaling reward) online programs can't reproduce that motivation.

Lectures are a very small part of the value proposition that universities make, and of course their functionality is already mostly duplicated by textbooks. Indeed, I'm not entirely sure why American universities primarily teach their students with lectures right now (versus books, video, Socratic method, etc.). I'd guess that the answer is some combination of tradition, students having a fundamental psychological preference for live teaching, and society in general preferring to believe that colleges are selling "education" rather than signaling, motivation, and lifestyle. In particular, I don't at all believe that the dominance of live lecture is driven primarily by a lack of good video materials. And availability of good video materials is the only thing that I see Coursera and Udacity really changing.

So: I'm not sure if there will still be state school professorships in 20 years, but I wouldn't bet against it.

(Note: for the record, I'm a fan of Coursera and Udacity. I also enjoy universities as they are today, though I wouldn't be heartbroken if they changed dramatically.)