As an independent researcher in machine vision, I take strong issue with your first statement - the gateways to science controlled by a beauty contest/club membership is fundamentally horrible science. However with your third paragraph you completely redeemed yourself - I'd argue if you've published, I'd guess you'd be quite happy with anyone reading the paper, it's just the emails from cranks that are a pain :-)
Practically I find that any paper I want can be found by tracking down the authors institutional home page - they always seem to find some way to make it available :-) Yay them!
The original idea of the internet as a highly distributed fault tolerant system that could route around localized failures. I will admit that part of that philosophy emerged when leaving arpanet and trying to get stupid bang email addys and netnews flowing over dialup :-)
L5 is L5 - if the system can't tell, then it ain't L5 - and the driver 'connecting' won't work - it's basically 'please doze for three hours and then deal with a crisis within the blink of an eye' Pilots have enough trouble with this and they have 6DOF in a mostly isolated space - I do believe, but all my fun comes from learning about the spectacular levels of failure and hubris in much smaller systems :-) Think of it this way - if it's AZ, OK, I pick Flagstaff when the snows have just started to come and grease the road
As for Elon being one of the driving forces behind the transition to electric - Rock on Elon, nice damn job - as far as L5 AV - umm, you're so full of it you need to drag a septic tank - I'm just a little fella, and I do admire the primarily visual approach, but this all really boils down to 'You look, but you do not see' - I've been in an out of machine vision for a long time, and the one thing that's always been true is that we fail, and fail again, when we ignore the gigantic cognitive stack that is processing that visual data and making decisions based on it (and if we don't ignore it, we are forced to admit we can't handle it) . L1,L2, L3... yeah, sure, it's not that hard -I think 'Logan' got the freight trucks in 10 years spot on. L5 - I'll believe it when Elon and the Times author leave NYC for a late fall drive to check out a VT ski home, say about mid October - storm gone thru, cloudy/foggy night, poor visibility, and the only brains you got are in that car - better hope it knows that leaves get ripped down then and piled up in the twists and turns of backcountry VT roads - my guess is there will be a funeral, or it's gonna be a multi-day trip - the ability of a machine to go 'I have seen this before' has nothing at all to do with 'I understand what I am seeing and can draw rational inferences' My 2c, admittedly playing with midget systems, but the failure modes are pretty much the same
Another upvote - Hello Fresh for exactly those reasons - I know how to cook, I like to cook, but it comes with a lot of baggage - this lets me focus on the parts I like :-)
I do recall from days past that some of the content came of the Keiwit games scodelib :-) Too bad Bridge never got ported, that game smacked the crap out of my Dad :-)
First time I met him was at an OOPSLA conference when C++ was just starting to get traction, while Smalltalk80 still ruled - so at the end, he's sagged back against a pole and we're all excited, asking if this ain't just grand - his response was priceless - 'It's terrible. Damn thing is too small to get any money for and too big to kill."
Zactly - Elon has been leading the charge to 'fear the killer AI' and has also made the claim that Tesla will have L5 autonomy in two years - for a car not only with the problems you point out, but a laughably poor ability to do the high school geometry to handle locating recharge stations - at this point the biggest danger from AI is 'we bet the company on AI and now the companies dead' I'm in and out of machine vision, so its not like I don't want to see progress made.
Hmmm - one serious thought - 'makers' thinking of playing with 1+ watts should also seek out videos of what happens to normal consumer and better non laser grade optics when used as focusing or steering elements - I knew of a lab that basically had very fine powered optical glass blown into the nearby walls (it was empty during the test) - and one less serious thought, from the same lab where I first encountered this sign... "Do not look in the laser with your good eye" :-)
I would argue it's not a one size problem - I use Fira for machine learning heavy code, which is primarily math, and ligatures for the various multicharacter operators make it more readable. Judge me, fine. But judge me in context please :-)
Been awhile since I used MM, but I do remember this - personally, I like using Fira with my code, but that's because there's a disgusting amount of math in it and it makes the reading a lot easier - your implicit separation between representation and presentation is spot on IMO
By that logic the gamma knife that has saved many people with cancer would not exist - before practical applications comes laying the basic foundation, and once that foundation is down, the genie is out of the bottle - I'm working on an autonomous drone, I am of course deeply opposed to weaponizing what I know is a ridiculously stupid system, but I know that it will happen - implication is not action, and knowledge is mostly neutral, it's the application of that knowledge that might be problematic - whichever way you slice it, it's a really complex issue, and anything worth having is often capable of both great and terrible things, depending on the application. AI is way too primitive to be malicious on it's own and will remain that way for years - now, as for abhorrent malicious uses of what we have now, here's a great example -- https://goo.gl/HmBY9k
Question - do you see any potential that horizontal versus vertical (with respect to system, not just euclidian space) has to do with cognitive stacking ? For example, is it low level feature detectors in initial visual processing that cause us to jump because they fire a 'snake' stimuli and then higher levels of the cognitive stack catch up and fire 'sorry, my bad' ?
I once attended some talk or other and that question came up - the answer given was flippant, but I always thought it was oddly perceptive...... "Dreaming"
Facebook is capable of monetizing current ML capabilities to do things like constantly tune individual news feeds so that its profit is maximized. Certainly counterproductive to some degree for modern society, but it is profitable. Straightforward goal. straightforward outcome.
Elon's claimed that Tesla will have level 5 autonomous vehicles in 2 years. Yeah, in your dreams Elon, that's a massive cognitive stack you can't fit in the vehicle, one we've never built out to such a degree, and your car's autonomy level is going to be a function of its bandwidth to the cloud anyway.
The mean part of me wonders if Elon plans on covering this wild assertion with 'Well, we can't now. Regulations and stuff' - if he's just worried about putting a weapon on a solar powered drone that patrols some area and shoots baddies, yeah me too. But that's just human stupidity, it ain't AI, it ain't intelligent, it's just a dumb program with effectors. All ML really is these days is real big boundary relaxation systems, calculating the massive number of coefficients we need to make the equations work out the way we want. But come on, it's stupider than a rock. It's an equation solver and its brittle as hell. As a tool I love what we can now accomplish, but there's a long road from here to Hal. It's not bloody magic, it's just math.
A key Chinese collaborator of mine on this https://hackaday.io/project/21966-quamera and I have chatted about the 'old' China and the 'new' China - he thought up, designed, and built a company around making a general purpose FPGA that allows for a wide range of imagers to be attached via standardized I2C/SPI or USB bussing. He's very much the new China, and while there are cultural differences, there's no question that an ever growing number can be either incredible collaborators or fearsome competitors - a lot of that choice is up to us.
Desperately sought just an extra 4K of RAM to see if a LISP expert system would get through a diagnosis on a Huge Aircrash Firefinder maintenance guide - had a kernel license, dug around and found a magic flag for a 4K block - tested it, seemed OK, put it out in the field, and the first time it ran, it grabbed that extra 4K and was instantly rewarded with a "Panic: out of swap space" and the whole damn thing dropped dead :-(
One thing I believe that's not being properly considered is the fact this is really a multilevel problem - if you want to make a smart IoT device you still want a reliable device. "On top of" is absolutely the best concept. Yes, the biggest computational bang for the buck comes from stuffing linux or something similar in, but who doesn't react with a smile when they see a multi year uptime reported - it's not rare, but it's not anywhere near a majority either - Start with the basic circuit, on top of that you add microcontroller logic to mediate between the higher level ARM SoC or whatnot and the base circuit, and just as importantly, you've also got a microcontroller based watch dog that will kick linux if it falls over. My experience in building this https://hackaday.io/project/21966-quamera is to never stop asking "Quis custodiet ipsos custodes" (anybody proffering analog solutions will be ignored, analog is hard and I am stupid)
Lot of good comments here - CV is heading in that direction, but in addition to the other mentioned problems, the needed data bandwidth rises exponentially - my own take is that we go broad in some ML work where we should go vertical (I believe that the same thing that makes you jump when you see a (non) snake is what your brain is using for image fusion feature detection. However, the real problem not mentioned is parallax - we understand some bits of it, but our brain does a lot of high level manipulation of the data - in the real world, it's two completely separate incidence angles for 2+ visual sensors seeing a point at a given angular relationship, but the same point is not the same thing - take an index card, paint one side red and the other green, and look at it edge on far enough away both eyes see each side, and its close enough for parallax - sometimes produces interesting illusions :-) When you get to complex 3D structures and parallax, the problem gets a lot nastier, because after all the work to get the data you want, you are then forced to throw offending bits of it away and start making educated guesses - my own take is "You do not see with your eyes, you see with your mind" :-)
Thanks for posting that - nice add to my collection, and I'd argue they work out some of the math far more than others (it's implied in other places, but its nice they actually worked it out instead of another bloody exercise for the reader) :-)