My favorite has always been Serpinski curve - the math behind Philip Jose Farmer's Riverworld :-) Here, you can print one yourself :-) https://www.thingiverse.com/thing:622627
There's a lot of history to it - the Parallel Data Processing Group (PDP) in the '80s was at UCSD - the real problem is that we don't have curriculums in place outside the top tier institutions that lead to research success, at least at the rate they used to :-(
Actually I like drones - my point is that no matter how smart it appears, software's still a pretty simple system doing what you told it to do. With what May wants, she will hatch a world where she can put you anywhere (video + ML to implant you), or hell just post online in your id, she can sign it and make it legal in a lot jurisdictions (because she poked a big fat hole in the crypto we use) and then she can use it as a big club on you - she can also control what you see on the internet, meaning she can easily make sure you see only what she wants you to see - and how long do you think that set of tricks really stays secret with 'her' - my point is it's hard enough to get this right, I think at the end of the day Arthur C. Clarke once again had the essence of it - let the 'smart' machine (I rate the internet as a very smart machine) do it's job, keep that job straightforward, the side effects are potentially horrendous.
This is horrendous - the entire point of the Internet is that it doesn't care, it's an inter-network, and it's only goal is to keep the packets flowing, at least when our dear leaders STFU. As someone in and out of machine learning - here's an example - I can litter the downtown with cameras, feed them into an ML cluster and if a child wanders off, it's mere seconds to find them again. Or, as in the great state of Connecticut, I can be so goddamn deranged that I want to put guns on drones, and believe semi-autonomous is a fine opening position. The problem is no matter how impressive the feat in ML, the systems involved have zero understanding of the outcome as a value judgement and they aren't going to gain one for a very long time. Do not teach machines to kill because they have absolutely no understanding of the difference between a beneficial act and a hostile act, they just reduce the error between where they're at and where you told them to be at. Going back to the internet you do not ever ever do this, because then what you are creating is a global censor. Happily for the moment such efforts are both localized and porous but this gaggle of idiots shrieking about terrorists and pedophiles need to SFTU and learn something, before they hatch something much worse.
I sometimes wonder if Tesla should be getting into autonomous nav with some of the evident problems they have with high school geometry and recharge planning. That said, the author of the article used SG&A expenses as a club to smack them with, without bothering to examine the reeds that make up the club. Tesla has more in common with Amazon than GM for a business model, they've got a voracious R&D appetite, and so on. They may well fail, but this article doesn't even seem to have a grasp on what they are, hence the measurements are all of time in furlongs per fortnight :-)
Wow - that actually makes a hell of a lot of sense - I thought that invalidated a startup :-) I like the combination of interests, I would believe if my homeowners ins co is running my monitoring and alarm systems, they do have a vested interest to try and do a good job. Good luck to you!
First I laughed, and then I thought this wasn't such a bad idea - sure you want to absorb the mind numbing true cost of the entire stack, then knock yourself out, be as 'Gollumy' as you like - get caught leveraging OSS for that and you pay the difference - a pipe dream, but a nice one - the real solution is more along - 'Dear John Deere, try not to be a complete dick all the time' but that rarely works
Why google keeps pushing Tango is an excellent question - there are some good people in the group, but in general the performance and achievements have been abysmal. Considering that TensorFlow and Tango come from the same organization would produce serious cognitive difference if teh googz scale wasn't taken into account - yes, I worked with Tango for a little over a year, and then gave it to my son as a game platform - seems the future is heading away from structured light to pure visual mechanisms, possibly with LIDAR or ToF for a little analytical support
Actually, not true - highly probable, yes, but not true - opencv already supports everything necessary to reproject a synthetic image based on a set of source images. If I'm taking a picture of someone directly in front of me and they're also visible to two cameras to my left and right in the same visual plane as I am the solution is trivial - after that it does get to be a question of how many camera, how much in phase, and your taste for chewing up gazillions of machine cycles.
Purely by inference based on ML/MV interactions with various bits of teh Googz, it will depend very much who is asking these questions - the closer you get to the tensor processing unit and its enclosing code, the more you're going to have to really know your stuff - further away in some of the ex-moon shots struggling to survive, I'd personally be surprised if the staff runs a reliable enough A* alg in their own heads to actually get to the room you're seated in - in short, googz is big - closer you get to the core the more hard core I would expect it to be, conversely as you move to the fringes, well that might be what you get :-)
Sigma and power aren't for loops, but they do represent the melding of iteration and operation - the key is your definition of the sigma above - silly computers are always trying to 'do' things, and often math can simplify problems too big to 'do' by fiddling with representation - so I do agree with you, I just think you're being cruel to the for loop example :-)
As a person who used to make their living completely off the msft stack and now is completely off it - I'd redo that headline to say every action of Ballmer's probably helped break his relationship with Gates - yes, Gate's Microsoft was very predatory and could be a very bad actor, but Ballmer's entire achievement was to help drive non-msft solutions to dominance and leave Nadella in a Sisyphean role -
Programming is pretty broad, as others have observed - in part of my world, scene reconstruction for machine vision tends to drag in tensor calculus, but the average enterprise developer is safe with the basics - also, good libraries mean you have the ability to use functions and processes you need, as opposed to getting recurrent nightmares about numerical stability when you have to code the realization of a complex process yourself
I do hardware engineering because I started my career when it and heavy math were required to get into software courses (and you first computer came as a bag of parts) - the camera itself is spherical build, wedging 26 cameras, 4 beagleboneblacks, a netswitch,IMUs,GPS, and batteries for the souls sucking multiamp drain into a 5" ball (OK, I gave up on the batteries for the moment :-) My primary "actual engineering" has been the control distribution circuits across 8 paired I2C/SPI busses at 4-8MHz and trying not to corrupt the bits in the process (a little bit of whining on Wolfram Alpha with the theoreticals, a lot of being very neat and tidy with the layout - the power distro itself with its various internal taps because power is analog and analog is proof that God hates us, the various passives and actives involved in everybody actually powering up in some more or less controlled fashion, etc - technically for the camera I use the Arducam FPGA against OV2640's and OV5640s, cause that gives me a nice (slower) pure I2C/SPI interface - the real goal is to feed the tensorflow stuff in the cloud, the camera geometry, imu, and gps provide enough correlations for the system to lock onto pretty easily, i.e. stationary spherical image reconstruction and spatial analysis is a no brainer (relatively speaking :-) ), and the back end system is now getting more betterer at using the IMU data stream as a primary assist when the camera is in motion (oh, and then there was the pesky little issue of delivering the GPS clock to 4 linux based computers to establish a common timebase cause we all know that linux is an awesome RTOS)
TL;DR; I'm an impoverished small scale serial arsonist of expensive components remembering that software means never having to leave physical proof of your boo boos, most of the time. OK, I'm actually a software engineer back into hardware engineering after 20 some years cause I need to build custom cameras for machine vision projects. Major recovered sills are CAD (Eagle), more geometry (layout), exercising credit cards due to I'm still bad a spice, lazily using standard cap and resistor values even when they're wrong for >this< circuit, and excessive shining about stray capacitance - since I need to work at 3.3V and I'm in my 50's, I can't see the parts so well and my motor skills aren't what they once were, so thats even more expensive crap I need (rework station, reflow oven)
I'm surprised no one has mentioned this yet - The Smalltalk environment itself is an excellent example of non-trivial Smalltalk - hey, you can even go in, fiddle with the language parser, and watch your whole world implode :-)