The Iron Yuppie

Just seven (!!) years ago, I wrote a blog post talking about the progress in computing power right around the time that Moore's law appeared to be ending saying that the change in processing power (and processing power/$) from 2001 -> 2012 was unprecedented and probably never going to happen again. 

TL;DR 2315x growth in efficiency/power in 10 years FOR LESS COST is pretty great.

Well, we're about 8 years beyond that, and it's... still pretty true! In 2012, top performance was 17 Petaflops/sec, and today, we're at 148 Petaflops, an incredibly modest 8.7x increase over 8 years. Boo hoo.

ON THE OTHER HAND, Google's public numbers around TPU pods (100+ Petaflops in January of 2020) are about $400k for a three year commitment, which is about $4,000 per Petaflops/sec. This compares with the 148 Petaflops/sec for about $200,000,000 for the top super computer, giving you about $1.3M per Petaflops/sec.

This compares with $5.7M per Petaflops/sec in 2012, and $14B ($14,000 M) per Petaflops/sec in 2001.

So, while the top end hasn't quite kept pace - it has been getting MUCH MUCH MUCH cheaper. Specifically, 1425x cheaper for the same performance in 8 years. 

CRAZY!

(Yes, I know that comparing generalized compute like in a super computer with specific performance in a ML accelerator isn't really apples and oranges ... but it's as close as we can get!)

Ok, I am the least qualified person in the world to comment on the current "I can't say what I want without no repercussions" mess, but I'm going to anyway.

First - that's part of the problem! What the hell people, everything you say isn't some miraculous gift from God that needs to be heard by everyone.

YOU'RE. NOT. THAT. FUCKING. SMART. (me included)

Can you imagine the kind of arrogance required to think "This may annoy a BUNCH of people but I better say it or the world will be a worse off. It's my responsibility to piss people off for the good of humanity!"?

Second - all these ideas that you're talking about are INCREDIBLY COMPLEX. They requires years of research, historical analysis, nuanced conversations, etc etc.

Suffice it to say if you think you've got the solution to homelessness, or global warming, or hiring bias, or anything because you had a dinner conversation, you're probably missing something.

"How hard could it be to create a global distributed data store. I can figure it out and I don't even have a CS degree." You feel that bile rising up in you? THAT'S WHAT IT FEELS LIKE FOR EVERYONE ELSE WHEN YOU SAY SHIT.

Third - I have yet to hear of a single issue that at any point in history was both proven correct and required being horrible/reductive to large groups of (often vulnerable) humans. Feel free to prove me wrong.

What about evolution? Heliocentrism? World is flat? In every one of those cases, the people forwarding those opinions had REAMS of data and hundreds of supporters.

The best, most recent, example I know of is Barry Marshall, who won the nobel in 2005 for proving that a bacteria caused ulcers, not stress. 

He was PILLORIED by the community FOR YEARS. But, let's be clear - his theory wasn't "This race are naturally criminals." It was that a foreign organism caused ulcers (which overturned a lot of research).

I honestly can't think of ANYTHING that started with "The homeless should eat each other for food" and turned into anything good. Again, interested in a counter example.

Before you add, YES I've read Galileo's Middle Finger, and YES, horrible scientists can be horrible people too. Nothing in that book contradicts this point.

My net take: Perhaps you should stick with what you know? If you really do have a FUCKING PHENOMENAL idea, that's awesome! Do some research, find someone in the field and chat about it, and get them as a supporter.

And do it again. And do it again. And again. And read this essay on how to be humble about your eventual published article/opinion.

And for those who say "But I'm just voicing my opinion on Twitter! It's free speech!" You're damn right! And it's other people's free speech to call you a naive tool!

Unless you are really confident in your opinion, and ready to stand up to slings and arrows, perhaps you should have a 1:1 conversation rather than blast it to 100s of thousands of Twitter followers?

Particularly if it's something outside of your core area where you may not have context? "How hard is it to make a profitable company? Just spend less than you make!" FEEL THAT BILE. REMEMBER IT.

But, AGAIN, this goes back to step 1. There is almost no way your idea is that valuable. If you really think it is, do the hard work and confirm it.

I find myself obsessed with a lot of the concepts that Wait But Why has been covering recently, but one that got me really thinking about is our immortality. To quote a section in a piece about AI:

Kurzweil talks about intelligent wifi-connected nanobots in the bloodstream who could perform countless tasks for human health, including routinely repairing or replacing worn down cells in any part of the body. If perfected, this process (or a far smarter one ASI would come up with) wouldn’t just keep the body healthy, it could reverse aging. The difference between a 60-year-old’s body and a 30-year-old’s body is just a bunch of physical things that could be altered if we had the technology.

• Your car lasts 100k miles, with fairly minor updates (oil changes, hose replacements, break repairs, but no significant damage to the engine)
• Your car has to go through highly varied environments (cold, hot, rocky, smooth, etc)
• Your car travels, on average, 30 mph (this is an average, so all your high way driving is evened out with your street driving)
• Your engine goes at an average of 4k RPMs
• You drive 30 minutes per day, and 10k miles per year

I don't think anything is too far off about this. By that logic, your wheels have rotated 69,000,000 times by the time you're done, and your engine has gone through 438,000,000 rotations. That's some pretty amazing engineering for something that is effectively unchanged tech in the past 50 years!

Given that our hearts beat about 1 billon times in our lifetime, it seems practically impossible that we wouldn't be able to out engineer that with the technology we have at our finger tips today, let alone in 10-20 years.

Everyone appears to be rattling sabers about AI being the next big threat (where "Everyone" = Bill Gates & Elon Musk, and a bunch of other people whose names are not generally well known). I think they're way smarter than me, and probably have nothing to add to that particular conversion.

However, one thing I caught today that I loved was the comment from the Netflix head of content that "The Netflix Engine" is "70% tech and 30% human." This is exactly right! I think that machines will (for some time, anyway) really struggle with that last mile and not falling into the "uncanny valley". This is exactly where humans can excel! Imagine you have a car production line, and need to finish putting in the very fine gear shift handle. You can either build a robot that has the sensitivity to not break an egg, but also handle spot welding the engine block, or you can focus the robot on the hard stuff, and give a human being the last little bit. I think we'll see humans continue to manage this last mile for a long time; it's so much easier than handling the 1% case with a general purpose bot. 

The news has been everywhere that Sony is LITERALLY worse than Hitler that they have canceled "The Interview." Hold your horses there, Hollywood elite.

First, I love Rogan and Franco, and would probably see them in anything. Now that I'm an old man, it reminds me of a simpler time - a time when smoking weed and being an idiot was consequence free. Well, technically it still is, but I'm going to pretend it isn't in case my kids stumble on this at some point in the future and need guidance about living life (NOTE: If you are in some way related to me, please ignore the previous paragraph). So having less of movies starring these folks is generally not a good thing.

Second, I think it's ridiculous to pull a movie based on one post to PasteBin. It's an anonymous network, is this all it's going to take to cancel movies going forward? What happens if someone posts to PasteBin they hate Michael Bay, will that be enough?

But, third, and most importantly, much as I don't like it, Sony is probably making the exact right move here. Opening a movie is one of the most fragile things you can do in business; virtually anything can throw it off - rain, snow, parades, holidays, non-holidays, who knows. There's just a million different ways for it to go wrong - that's why studios fight over big weekends, and make sure it opens on more than 3,000 theaters, and have huge week-long press junkets leading up to the release to make sure there is as small a chance as possible that you won't show up on the first day.

Even worse, you could have a great opening, but some surprise comes along and knocks you to #2 - your movie is basically worthless at that point. Movies have a HUGE drop off (even great movies) the second week and forward, and, while there are always exceptions, it is basically a 1 in 100 chance that you'll see anywhere NEAR your opening weekend audience in week two, let alone going forward. Because you're now a #2 movie, that means you don't get to be the tentpole when it's time to sell the entire bundle of movies to Netflix/HBO/USA/whatever. And, at this point, not being #1 is virtually guaranteed when you factor in movie theater chains are starting to pull showings ("Let's see, we can show Annie, which is totally neutral, or show The Interview, which, while it got a bunch of news, which would be good, but has an ever so small chance of scaring people  to not coming. No brainer.").

On top of that, let's say Sony opens the movie, and there is a 1 in 100,000 chance that someone does something crazy and kills one person in a movie theater on the other side of the world after the movie opens. At that point, Sony Pictures is basically over. It's just too crazy to even think about, and risking an entire studio over one $42M picture is a cost-benefit ratio that is too big to handle.

So, with all this lining up against the movie, they pull it. It's not awesome to have a miss, especially on Christmas, but it's built into the studio system, and Sony knows how to survive that.

The far more concerning thing is that it appears that unrelated studios are now pulling ancillary products, like Team America: World Police or a Steve Carell movie set in North Korea. That's really worrisome - does that mean Seth Rogan and James Franco are now done making movies? Hollywood is a very skittish sort; they're going to give as wide a berth as possible just to make sure nothing bad ever happens. Not that there's some huge demand for North Korean movies, I just think this is beyond stupid.

BuzzFeed has had an absolutely amazing run, leading up to a $50M investment by Andressen, Horowitz, the hottest VC firm in the world. However, Frédéric Filloux of Monday Note is decidedly less enthused. A key point in his open letter is below: 

Fact is, quality content does exist in BuzzFeed (an example here), but in the same way as a trash can contains leftovers of good food: you must go deep to find it. 

Well, trash (one assumes) is stuff no one wants. A listicle of "17 Ways the 90s Were the Best Decade" is wanted, whether you like it or not. But going beyond that, why the hate for listicles at all?

I have two theories - either it's a distaste for ads, or a distaste for the medium. To evaluate both these ideas, let me take you to a simpler time - the year is 2000, the year that the newspaper business had the highest revenue in its history, and the Pittsburgh Post Gazette (a paper with over 175k subscribers) has its 36 pages broken down in the following way:

• 10 pages of majority news (though 4 of these are reprints of stock quotes)
• 16 pages where news makes up 50% or less of the content (with the remainder being ads)
• 8 pages of classified ads
• 2 pages of comics

Perhaps that's too recent for you? How about September 10, 1914 - World War I is well underway, and here's a copy of the Times Dispatch:

Of the 14 pages in the newspaper, it appears that it is made up of:

• 1 page of pure news
• 11 pages of between 40%-65% news (with the rest ads)
• 2 pages of pure ads (classified)
• Three quarter page comic strips

There's quite a bit more stuff in there that's not exactly news either - my favorite is a 100 year old native ad!

By contrast, this listicle of 26 Animals From Around the World has just 1 300x250 ad, and one native ad (at the bottom). That's practically pristine!

Is the issue the fact that Buzzfeed repackages other content and draws in users? I'm not a huge fan, but that isn't exactly the end of the world either. It's all free (e.g. the reddit content they repackage is just as free as their content is) so if you don't like it, go to the source - where hundreds of links are scattered between countless pages and make it nearly impossible to find it all.

Or perhaps the issue is that news isn't 100% of the paper at all. But that can't be the case, since news has ALWAYS been advertising funded as evidenced by the examples I listed earlier.

An interesting alternative to just bad mouthing listicles is to think about news organizations as having always been made up of two parts - the news, and the other stuff that funds the news. In newspapers, this was made up of classifieds and ads; on the web, it's listicles and ... ads. There may be variations of this equation in the future (how much time and real estate a publication can/will spend on each),  but it seems highly unlikely there will be too many sites that have the former without having at least some of the latter.

(I don't know if this needs saying, but the following post has absolutely no information contained therein that I learned while working at Amazon. Also, I have a position in and am long on AMZN).

You know what sucks? Margins in retail. 

1. Amazon makes the vast majority of its goods selling online today. But, 85-90% of total purchases occur in the physical world. Amazon wants a piece of that.
2. Amazon is all about selection. They're "Earth's Biggest Selection". Can't have selection without every single product in the world - online and off.
3. Most importantly, Amazon has the opportunity to close the loop - to keep money flowing through its system, and never let it out. 

The third point would be absolutely transformative. Think about it this way - today, I go to Target and buy some sheets. Target goes and buys some floor cleaner. The floor cleaner company pays their employees. The employee goes and buys a Nintendo DS at Best Buy. 

Everyone of those transactions require a percent off the top for credit card, debit and ACH fees (on average about 2%, similar to low end retail margins). 

Now imagine that Amazon is the player at each step - they sell the sheets, floor cleaner, gift cards (or creates a "store your money with Amazon"-style bank), and Nintendo DS. Without doing a single thing they don't do already today, Amazon has participated in 4 transactions with zero processing fees. Instantly they've doubled their margin.

Amazon could drop their prices and demolish retail, or keep their prices the same and vastly improve their cash flow. Regardless of their choice, it's such a huge strategic win, I'm not sure why everyone is focused on all the other points. Keep the money and you win by default.

There's a spirited discussion over on the excellent Wait But Why on Why Americans Don't Love Soccer.

I think the majority of the items are spot on correct - they're some really basic items that just seem like no-brainers. They would not affect the flow of the game at all, and improve the fairness (reducing things like incentives for flopping and fake injuries). However, they missed a big one that I think would be even better - more refs.

If you look at the other major sports - Football, Basketball, Hockey and Baseball - they all have much higher square-footage to ref ratio. For example:

• Football:  110m x 48.7m = 5,363 m² / 7 refs = 766 m² / ref
• Basketball: 28.65m x 15.24m = 436.6 m² / 3 refs = 145.5 m² / ref
• Hockey: 61m x 30.5m = 1,860.5 m² / 4 refs = 465 m² / ref
• Baseball: Varies, but around 100k ft² = 9,290.3 m² / 4 umps = 2,322 m² / ump
• Soccer: 105m x 68m = 7,140 m² / 3 refs = 2,380 m² / ref

Soccer has the MOST square footage necessary to cover for each ref. And, if you change baseball to just looking at the infield (where most of the action requiring umpire evaluation happens anyway), soccer has a practically 4x greater difference than the nearest other sport. Yet fouls in soccer are just as serious (or worse) than other sports. 

More refs would massively improve the game - diving would be less likely and more bad behavior would be caught and punished. I love the idea of changing the rules (particularly the two tier penalty in the box to prevent cheap dives to get penalties), but let's start with just more accurate calling of the game.

If you have a position, back it by data. If you get new data that contradicts your existing position, either fix your data or change your position.

If you can prove I'm wrong, I will change my mind. I’ll spin on a fucking dime.

Science adjusts it's beliefs based on what's observed

Faith is the denial of observation so that Belief can be preserved.

This is just a fantastic article about why Gen Y is unhappy (despite the not-obviously-great title): Why Generation Y Yuppies Are Unhappy

The TL;DR is just that we (I think it's we, I might be on the border between X and Y), for our entire lives, have set our expectations high and, by definition, not everyone can be exceptional. Or, basically, this awesome equation:

The only bad part about the article is that there's effectively no way to break the cycle without a) a massive depression of expectations (in this generation or the next) or b) millions of people working to hard to be happy with what they've got. I don't see either happening any time soon.