231 karma · joined March 15, 2012
That's what may change if these new smart watches find their killer app. No one really needed a calculator, heart beat monitor, or other gimmick on his wrist, so digital watches never pulled away on features. That might change now.
Would I rather wear an intricate piece of jewelry or see why my phone vibrated without reaching into my pocket? It starts to be a different calculation.
Watch purists may go the way of vinyl purists, slowly dying out as the competing technology offers more and more convenience.
Whatever their merits as engineering marvels, expensive watches are Veblen goods that are worn to display wealth. Similar mechanisms can be put into pocket watches, but that's a much smaller market. Expensive watches have the appeal they do because they fall on the same continuum as everyone else's watch. Change the middle of the watch market and the current high end will look as absurd as a Vertu brick phone.
There's also a budgeting issue. A hotel spending hundreds of thousands of dollars installing wifi is going to want to recoup that cost directly instead of assigning it to overhead, and once they start collecting fees the revenue is hard to give up. A hostel can just add a $50 consumer router to their existing internet connection.
Then you have to write software that lets you both experiment and release your models to production. That involves writing pipeline architectures to apply things like feature extraction and pruning in a consistent way, and to make sure the result can be serialized and deployed. Off the shelf packages typically haven't solved these problems very well, so you have to make sure the thing that looks good in the scripting environment is reproducible on new data.
Then when you have a model and can deploy it, you start working on automating the training process so the model automatically adapts as new data comes in. Usually the customer has gotten the impression this was happening on day one, so you have to rush to deliver it.
Then, you deal with customer complaints that something that would have been obvious to a human is wrong even after they corrected it.
At some point you try to measure the gains you're offering over the non-ML system you replaced, and try to tweak those metrics until they make you look good.
If you're lucky, you got to experiment with some interesting algorithms somewhere in the middle, but you probably got the best results from something fairly standard like random forests and not the latent bayesian slice sampler you dreamed up when you first heard about the problem.
Good advice. So why do you give the same amount of attention to variable naming and early exits as to much higher level issues? I don't think the difference between Carmack and a random developer has much to do with adherence to coding standards.
Taxes are always computed on net winnings annually. Filing as a pro can affect the rate and allow deductions for expenses, but even amateurs can deduct losing sessions against winners if they have documentation.