Monday
Jun032019

Apple's New 2019 Mac Pro isn't for Professionals, It's for the Luxury Market

Apple sometimes has a way of paying a lot of attention to the wrong things.  UI, Users, Accessibility - maybe the right thing.  Professionals, maybe the right thing.  The thousand people that can afford a solid gold apple watch... um... sure the profit margin is fantastic but maybe they are not the reason apple is frequently wonderful.  Lust worthy technology, to a technologist - or perhaps just me, isn't going to really be about using some beautify soft inferior metal that can't handle a picnic on on a boulder at the beach without looking like scrap.  Sorry gold.

I saw Apple's WWDC mac pro and had bummer moments aplenty.  

The case is thick.  That grill is deep.  And my first bummer moment - electronic discharge milling or convential CNC can cut precise half hemispheres into a 20mm slab of aluminium at considerable waste... but why go there at all with a design that reminds one of bee's and the breeding backs of the surinan toad?  What professional need is met with a insect nest friendly computing enviroment, a ban on workplace squirt guns?   A Unique solution to a difficult problem is quite a different thing than a silly solution to a well solved problem.   Posh, sure, useful... umm.  Useful like optional roller feet on a $6000 to $59,000+ desktop because are accidents never happen. Useful like a monitor with grill that lets those damn little beetles fly right inside your $6000 pro monitor when you are filming on set?

Then we move on to the Xeon CPU.  I like xeons... historically nice preformers solid choice... and the 14 nanometer 28 CPU version is available on New Egg for $10,000ish dollars.  A reminder, 14 nanometer is the best intel can do these days... but maybe they will soon have 10nanometer chips to be marketed as 7 nanometer despite microscopic evidence to the contrary.   Apple already has 7 nanometer components in the chips in the Iphone tennis... so awkward tech lust moment - its like falling in love with Juliet and trying to chase her down to go all "But soft what light" on her, but then getting waylaid by that mopey goth bitch Ophelia and being forced to hear all about everything rotten in Denmark and how drowning might be the only way to ease her sorrows so you know, when you finally do get away, that it's all going to be Act V Scene 3 for you and Juliet even if you both find each other?     

Really Apple, I have to buy ophelia?  Ophellia?  Even if I want Juliet?

Hook line and sinker.  Because next comes the GPU options.  And Nvidia's CUDA isn't in the offering, and the story is much the same.  

If you don't program AI, let me expain there's a training phase that takes hours or days to teach your computer dog to hold a biscuit on it's nose, followed by a trick phase that makes the biscuit vanish in a magical instant so quick that everyone thinks it must be easy - after all - obviously the dog did all the work.  If you ever trained a dog you might know better... and this is my beef with Apple's state of AI:  With apple you can have a trained dog easy enough, but to do any dog training is really going to be slow and painful in Apple's ecosystem.  For me... CUDA and machine learning go together.  It isn't that you can't do AI with just a single CPU or AMD boards or Apples modest Metal AI tools... it's just that running tensorflow on that 28 core xeon is never going to be fast compared with what you can do for less on a CUDA freindly graphics card with a thousand of cores.

...so.   Yes.  Apple made a AI development bet that looks stupid to me and hints a a fundamental forward technology disconnect.  The AI community probably isn't going to be flocking back to mac's soon.  And it kind of makes my brain hurt - apple should deeply embrace AI given its niche in the companion / personal assistant products - and instead... they made design decisions for their "professional" market that amount to "AI dev?  eh maybe next time?  Maybe never?  Oh, and good luck finding third party vendor and a hackintosh ktext and drivers so you can run Cuda boards in our walled garden."  And worse, that choice may likely keep Apple's Siri in the slow lane, compared to Amazon and Google, unless there are plenty of windows and linux boxes in Apple with cuda boards just exactly for deep learning and training.  But given the state of Siri, I think they don't exist at apple. 

Apple's new and different and moderately less useful AI developer tools -  isn't going do it apple.  Forcing devopers to  buy expensive GPU options that they almost certainly would want to immediately replace isn't super great for anyone but apple and hardware vendors.  It makes other options look pretty strong.  I really wish apple saw it's professionals as less monolithic and more in need of future proofing technology side bets  and able to support special use cases.  

I like apple.  I think I'm done with apple.  I don't need a specialized supercomputer workstation that isn't targeted to my needs, and more or less by random decreem can not be targeted to my needs.  I hate the pain of switching, but it seems like soon enough, that's me.  I'm not part of the luxury market, or whatever niche film maker video edit suite pro market this Mac Cheese Grater is tailored to (8K video hardly exists, but if that's the 2-3 year future of film, maybe they are the targets of the quirky engine tuning the 2019 mac pro has?)  

Beautiful, prideful, sleek, powerful and able to roll off a desk onto concrete floors in less than a second?  sure, the 2019 Mac Pro is all that.  Something I want to own?  No, not for me.

Monday
May202019

Game of Thrones Should Issue Season 8 Shorts

Game of thrones, maybe I never got you.  Shows done, and what are the fans left with - a hot mess?

I firmly believe HBO should seed a little "OH HOLY SHIT" good will by giving Bran-esque glimpses of the future... at least half of which involve political assassinations if Bran raises taxes to restore Kings Landing or should he fail to do so.  Enough wiggle room that the viewer is left to wonder what comes to pass.

 Dinkle getting a drop with a sword... Massie with a mask... Bran falling off a horse or tumbling down a tower... daenerys as dead zombie nom nom or maybe resurected by a lucky drunk.. Jon Snow with the night watch over taxation..  A diplomat from disputed lands... 

yea.  No reason any of it needs to be anything but one of bram's fever dreams... 

Saturday
May182019

Lee Maze, a new rust crate on crates.io

Leemaze is my latest crate, offered as a MIT/Apache2 licensed, open source gift to the Rust programming community.  It's named after the Lee Algorithm https://en.wikipedia.org/wiki/Lee_algorithm.  It solves 2d, 3d, 4d and 5d block mazes.  A 5d block maze is basically your morning commute with moving walls (cars) and 2% driver crazies that cause ..predictably best avoided.. random daily events, so... yea... if you are trying to avoid things that haven't even happened yet you might be living in a 5d maze world.   

What does leemaze do?  It gives maze directions.  Feed it a boolean block maze and it will find one contender for shortest possible path.  Or maybe it returns a None().  Or if your entrance is your exit a empty vec![].  Any maze with a open 2x2 block may have multiple shortest path solutions.   

Why did I make leemaze?  I was learning the superb KISS3D rust graphics library in April and wrote a 3D block maze.  Then I decided to generalize my code - navigation is a common computing problem after all... and then realized that rust didn't have jump out at me obvious Lee Maze crate that did the same thing.

how does it work?  Forgive the unchecked code- i'm in a HTML editor but my intent is all goodnesss -

       use leemaze::{AllowedMoves2D, boolify_2d_maze, maze2dpath};

let mymaze = vec![

                   vec![0,1,0],

                   vec![0,1,0],

                   vec![0,0,0], ];

let maze = boolify_2d_maze[&mymaze, &0];  //convert type to boolean trues and falses,  

let northsoutheastwest = AllowedMoves2D {rules: vec![(0,-1),(0,1),(1,0),(-1,0)] };  //allowed x,y axis moves.  

let entrance = (0,0);

let mazeexit = (0,3);

let maze_escape_directions = maze2dpath(&maze, &northsoutheastwest, &entrance, &mazeexit);

And that's it, out comes a None() or a Some(contender_for_shortest_possible_maze_directions).  Um... are the directions indexed? YES, from the list of AllowedMoves2D you provide --> if your allowed moves were 'northsoutheastwest', maze_escape_directions might be vec![1,1,2,2,0,0] - for south south, east east, north north.  Use the directions to move your robotic death tank or NPC towards its goal.

Ok, all well and good but I want to do a 5d maze, hows that work?  Build a 5d maze and just replace AllowedMoves2D --> AllowedMoves5D,  boolify_2d_maze --> boolify_5d_maze, and maze2dpath --> maze5dpath.   

So a gift of code to the rust community, enjoy.  And good luck on your morning commutes.  :)     

Tuesday
May142019

A Round Up on why it sucks to be Bayer

A jury awarded a couple 9 billion dollars in damages for two cancers they believed related to the use of the herbicide roundup.  It pains me somewhat, but two human lives aren't worth 2 billion dollars, not according to various federal agencies put the value of a human life at somewhere between 2 million and 12 million dollars.  Drive off a road that was improperly built to tilt out (rather than into) the curve and your life is only worth 6 million.  Get poisoned because a business disposed of lead car batteries in your cistern well, and magically your life is now worth about 9 million plus multiple times property and collateral damage.  

Yes, a floating value on human life that depends on how special your death is ... curious.  But let us accept that some magic value per human life exists, round about 9 million.

So the couple in California, Alva and Alberta Pilliod, both came down with Non-Hodgkins lymphoma,  after a potentially misguided effort to keep lawns dandelion free.  Non-Hodgkins lymphoma has a 1 in 5000ish rate for singles, or about a 1 in 25 million-ish by chance alone odds for Alva and Alberta to both get the disease by bad luck.   Long odds, but the stats mean there should of been roughly 10 double diagnosis couple cases in the United States in 1973, the year before roundup's introduction. Their lawyers argue the primary cause of their cancers to be exposure to Roundup herbicide, and they have been awarded 2 billion dollars.   To me this seems a bit excessive by most measures.    

This is what 9 million times two looks like:  $18,000,000.  This is what 2 billion dollars looks like: $2,000,000,000.   See the three extra zeros?  ($18,000,000 : $2,000,000,000)  

If the jury accepts that Alva and Alberta (and presumably some of the other 14,000 pending court cases and anticipated 65,000 new diagnosises of Non-Hodgkins lymphoma each year) were harmed by roundup or, more plausibly, boogie man chemical impurities in roundup, why would they award enough to greatly limit payouts to other victims?  Bayer only makes 7.3 billion-ish in crop profits per year, so if each victim gets $1 billion for cancers weakly linked to herbicide use.. that be seven payouts before Bayer starts selling assets.

And, worst case senario by historical incidence rate and Roundup's 1974 introduction, I don't think Bayer is responsible for even one in ten of new cases of Non-hodgkins lymphoma.  Look at the data, I don't see a trend bump post 1974 in incidence or mortality.  A trend upward in risk for the elderly is present, largely as the population of the elderly grew between 1950 and 1990.  There is no off trend deviation after 1974, the year Roundup was introduced.  My conclusion: if roundup causes cancer at all, it's still way less dangerous than sunshine.  

[CANCER RESEARCH (SUPPL.) 52, 5432s-5440s, October I, 1992]. Non-Hodgkin's Lymphoma Time Trends: United States and International Data.  Susan S. Devesa2 and Thomas Fears. Epidemiology and Biostalistics Program, Division of Cancer Etiology, National Cancer Institute, Bethesda, Maryland 20892

Bayer (cough, monsanto, cough) helps keep food inexpensive and plentiful.  Bayer reduces world wide hunger and even heroically saves lives.  That said, its hard to be a fan of Bayer crop products - one can argue roundup kills beneficial wild plants,  their enviromentally persistant insecticides have encouraged bee colony collapse, reduced butterfly and moth species, damage animal immune systems in ways that encourage bat suceptibility to white nose fungus, kill fish, increased frog and amphedian fungal infections and maybe weakened some animal immune systems enough that they allowed some rare cancers to persist.

Insecticides kill bugs, herbicides kill plants, roundup is a herbicide, and they are necessary evils in the modern world.  Just take care with using necessary evil unnecessarily.  Maybe GM dandelion/lawn weeding/lawn mowing munch bunnies will be a thing someday, along with AI-assisted UV laser pest control that targets mosquitos, dandelion seeds and any stray much bunnies that happen to be getting a mosquito bites.?    

Marketing "necessary evil" to kill every harmless dandelion in every household lawn world wide might be where Bayer went wrong.

Thursday
Apr182019

Presently Indeed.com Lists 79 Positions wanted for Reality TV Show Writer

I don't know how many reality TV shows there are... but 79 open positions for a industry hints at a total of somewhere around 1500-1600 writers for those all those "reality TV" shows.  Google gave me a list of 50 reality TV shows so that puts the number of writers per reality TV show somewhere between 2 and 32. 

Why do they need so many writers for television that proports to be unscripted?