Posts by DrBobK 117 posts • joined 17 May 2011
See the incredibly beautiful study of Hecht, Shlaer, and Pirenne (1942) Energy, quanta and vision. J. Gen. Physiol. 20 819-840. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2142545/ Dark adapted human rod cells require between 5 and 8 quanta on average to produce a neural signal. After the effects of absorption by parts of the eye between the outside world and the retina this corresponds to an ability of the humans tested to detect between 54 and 148 quanta (i.e. that's all you need to see a light under optimal conditions - the integration time is about 100ms, so all the quanta have to arrive within 100ms for optimal performance).
Here is the relevant section of the paper's abstract:
"With these three corrections, the range of 54 to 148 quanta at the cornea becomes as an upper limit 5 to 14 quanta actually absorbed by the retinal rods. 3. This small number of quanta, in comparison with the large number of rods (500) involved, precludes any significant two quantum absorptions per rod, and means that in order to produce a visual effect, one quantum must be absorbed by each of 5 to 14 rods in the retina. 4. Because this number of individual events is so small, it may be derived from an independent statistical study of the relation between the intensity of a light flash and the frequency with which it is seen. Such experiments give values of 5 to 8 for the number of critical events involved at the threshold of vision."
Your truly,
A Vision Scientist.
I'm amazed at how many of the tin-foil hat brigade showed up for this one. If you are that paranoid about surveillance and the lengths higher powers will go to to get it, then surely you realise that the NSA have 'special' code inserted into all smartphone OSs and every network adapter driver for every operating system used by more than five people. Despite all this guff about the security of open-source, no-one actually checks everything (and they don't know the 'one special trick').
That's a great move. I like them. I (genuinely) only use TeamViewer for home stuff. At one point, not because of usage, but because I clicked a button to try some new feature for free, they decided I must be using it commercially. I emailed them. They checked their records. I was put back to 'free' and they even apologised. All also done very quickly. Quite amazing.
The reviewer said "At nearly £300, it's almost as much as a pair of Sony WH-1000XM3 headphones – which is one of this reviewer's daily drivers". These are *headphones*. They seem to be being confused with WF-1000XM3s in the comments. Those are Sony earbuds. I now don't know which I've missed out on buying for $280 from Orange-Idiot Land. One a related note (again risking bud and phone mixups) I had some Sennheiser Momentum Wireless Headphones which sounded great but were quite incapable of maintaining a connection to my phone when there were other bluetooth devices nearby - something which does occasionally happen in places like aeroplanes. I replaced them with Sony WH blah blah blah which might not have sounded quite as good, but did work wirelessly.
Can't someone just recreate the simplicity of Borland Builder as it was in the late 90's? Made coding the interface trivial so you could concentrate on the stuff behind. I'm not a professional programmer, but I program a lot (I'm an academic) and nothing seems to have replicated Builder's combination of simplifying building the interface, yet still letting you do whatever you want behind the scenes (I program experiments in visual perception, so I don't care about things like database integrity, but sometimes I want to write programs that other people - students - can use easily).
Isn't writing some code to normalise image stats and provide a gui to tag regions of pixels a 10 minute job in Matlab if you know its gui builder and have the image processing toolbox? I did something like this to allow someone to tag regions of a scene and generate a recolored version for eye-movement analysis a few years ago and it was trivial. My understanding is that it is this front end that gets your software classified as a munition or whatever, not using the tagged results in training some network. Absolutely crazy.
Do these need a wire to connect to a sound source? Despite being a bit of Senn fan (HD25 and HD26Pro are fantastic indestructible headphones) I have just sent back a pair of Sennheiser Momentum 3 wireless noise cancelling headphones because, although the sound quality, noise cancellation, and comfort, were great, the bluetooth performance was terrible - in an aeroplane where lots of other people were using wireless headphones, they kept cutting out horribly. I got a pair of sony wh1000-mx3 (such a romantic name) instead.
Out of genuine interest, what is currently wrong with iOS 13.2.3? In addition, has anything been done about MacOS Catalina yet or is it still a disaster zone? Whatever is/was wrong with iOS, seems minor compared to what I read about current MacOS...
Can't algorithmic programs learn by example? Isn't Doug Lent's Cyc (https://www.cyc.com/) algorithmic rather than an artificial neural net (so far as I can see the book is really about ANN-based AI and not AI as it was practiced in the days of LISP machines and 60's cognitive science)?
Partially it is down to IT service management policies where I work. If you use a windows machine for office stuff it ends up being centrally managed which mean you lose admin rights. If you have a Mac none of this happens. I used to work with Macs in the late 1980s and came to hate them, but I thought I'd try try giving them another go a few years ago because the admin rights business was so frustrating. The terrible instabilities of MacOS 7.5.3 etc. from days gone by were no more. Everything worked without a hitch. I could run enormous displays and use a nice keyboard at work (and even run a nice graphics card in an eGPU) through a dock and then take the Macbook home by just unplugging a single Thunderbolt 3 cable. I could have set all this up from Windows but I suspect it would have been harder to do. So, all in all, I don't think Macs are leaps ahead of Windows, but they are easy to use and once I'd made the switch I liked the experience.
Bollocks. Here I sit with an enormous Windows box with a rather good Nvidia GTX 2080 Ti running an HTC Vive Pro Eye that I program with C# in Unity and next to that some HP dual Xeon server running Linux in which I develop (not use, develop) multispectral physics-based raytracers for vision research. But I'm typing this on a MacBook Pro because that's what I like using an an office machine. So piss off with your "People who buy Apple computers are either not IT literate or idiots, no ifs buts or maybes". Idiot.
Is this the same OPPO who sell audio gear? Planar-magnetic headphones which are too expensive for me to buy (a rich friend has some though), a very good portable headphone amp which I do possess, and desktop headphone amp which is way out of my price range. All decidedly at the high (but not totally, totally insane) end of things. On that basis I'd expect them to be selling £1000+ phones with maybe one lesser option.
(The megaphone is the closest icon I could find for 'audio')
The current NN craze (CNNs, deep learning etc.) have pretty much nothing in common with the real thing in which the individual neurons have complicated time-dependent dynamics even before cyclic connectivity give the network itself complicated dynamics (difficult things which are intentionally avoided in artificial neural networks). Many years ago I was in Santa Fe at the Complex Systems institute where I met a mathematical physicist (ex-string theorist I think) who was trying to work out how the different modes of 'chewing' in the stomach of a lobster were produced by the 12 neurons in the somatogastric ganglion of the lobster - repeat just TWELVE neurons. He was part of a whole team of heavy duty applied mathematicians, they hadn't made a great deal of progress in understanding the 12 neuron system when I heard about it (but they did do experiments and got to eat their subjects once the experiments were over, which was good). It is relatively simple to construct a model that approximates the neural network in a nematode (although figuring out how much detail you need in the models of individual neurons to adequately mimic the behaviour of the real thing is tricky), but that is utterly trivial compared with understanding how the properties of network give rise to particular patterns of neural behaviour and understanding why these patterns of neural behaviour are of use to the organism as a whole.
The idea that everyone in the 50% of the school-leaver population who were encouraged to go to university would actually benefit from it is crazy. A lot of people were essentially conned into spending a lot of money on something they probably didn't enjoy (the learning bit, not the social life bit) and that, in truth, wasn't going to land them with better jobs than they could have got with a university education.
All this business about needing PGCEs or other teaching qualifications to be hired to teach in universities is either odd, weird, or wrong. These days in my hallowed institution (in the worldwide top-100 etc.) lecturers are hired regardless of whether they have any teaching qualifications (we do care about track record in research though) but have to do a teaching course during their first two years of employment. This happened some time after I started, so, as a member of the old guard, I have no teaching qualifications of any kind yet still get to be Prof. and teach at all levels. If, however, I retired from university and wanted to do some school teaching then I really would need a PGCE or something (although I think I could teach at a private school without any teaching qualifications).
Suppose I want to get an input n times and each time I get it I then test it's value and then choose what to go and do on the basis of this value before returning and getting the next input. How do I do this in a language like Bosque?
[This isn't some cynical trick Reg-comments-esqe question - I don't know the answer and I'd genuinely like someone to explain it to me.]
I have some audience tapes I made of the good old Grateful Dead in Paris in 1981 (I think - I was a touch out of it then and have a touch of grey now, so memory isn't that great). I'm waiting for the French authorities to pop over on the basis of their amazing powers of deduction etc. Is this seriously down to 'Blues for Allah'? Jesus these people are mad.
Wow. Good for IBM. I'm interested, as an academic who studies vision, in assessing the importance of potential low-level statistical texture descriptors in object or face recognition. I've just filled in the form asking for access to the database.
People who want to do evil things with coded images will already have done it if they are really any good at doing their evil things.
I wasn't noticing enormous differences (or being enormously impressed) until we got to the food pix of the full-English. The Apple one is pretty good. The other two (both Huawei?) are appallingly bad. Food pix seem a pretty good test of image quality because for the food to look appetising you need to capture colour variation, texture, and highlights accurately. I'd hazard these are also pretty important for other natural materials like skin. I'd be surprised if the phone that took the bad food pix could produce good close-ups of faces.
The 50x zoom pics are a joke (but I expect everyone with half a brain must have figured that one out). I can get fairly decent ridiculous telephoto images with a tripod and an 800mm lens on a 35mm full frame format digital body with decent lenses and a tripod (actually, for me, 400mm lens and 2x teleconverter - all Canon stuff with red lines around the lenses, all very old (e.g. the body is a 1Ds Mk 2 - less than £300), all very e-bay) but even then it is hard (I was doing this for work - it isn't a focal length I'd normally use for fun). I'm surprised you can get anything sensible hand-held from a phone (congrats for managing to at least get the centre of the postbox sign, if only to demonstrate the lousy image quality).
Had a VPN for ages for non-pron reasons. Mine costs a trivial amount (less than a pound a week - NordVPN) and was easy to set up and use. Surprises me that everyone doesn't have one. You can get local news from countries that block it otherwise, in reverse get UK-only sites when abroad, get US media services, and so on.
As someone who is considered by some people to be worth giving a fair whack of money to to dispose of on kit as I like here is my take on high end laptops. I like to have exactly the same work environment wherever I am. In the office I have some lovely 4K HDR screens. I can plug my laptop into these through a nice box (and also connect to a big bit of spinning rust for backups) and have a great desktop experience (also via a wireless keyboard and mouse). If I unplug my laptop for the road then the work environment is the same but now seen through a pretty (very) good laptop screen. Software, data, operating system, everything, is exactly the same. I don't have to worry about internet access or any of that hideous roaming profiles crap to ensure this. I now do this with one of those overpriced Macbook pros, which I prefer to Thinkpad X1s that I used to have. Whichever you choose, the laptop powerful enough to create a fab desktop experience and be well enough equipped to work well on its own is a great thing. It makes work reliable and consistent. I don't care whether I can upgrade it or not. When I want a new one I will get a new one.
I replaced my old X1 with a Macbook Pro (mainly for reasons to do with having to have a managed desktop and no admin rights on Windows, whereas free to do what I want with MacOS or Linux - already have Linux machines for rendering etc.) and I'm pretty happy. The keyboard on an X1 is undeniably better, but the distance between the keyboard and the front edge of the case was painful for me. I did stick with Lenovo a bit though - got the first version of the Thunderbolt 3 dock, which is an excellent thing - from laptop to dual monitor desktop with multiple external disks, by plugging in a single cable.
My great great grandfather was an engineer. He emigrated to Australia along with the steam engine he was employed (for the rest of his working life) to keep running. Being an engineer didn't used to mean being able to design a steam engine from scratch, it used to mean being able to keep it going. I'm heavily indebted to BG engineers who can keep my central heating going. Something that, as a scientist, is of course, completely beyond me.
We need a reference for the flatworm claim. I've taught the flatworm story for years. The 'memory' is just stress hormones (generated when the 'donor' flatworm has to learn a stressful task) which affect the rate at which regenerated (or cannibal) flatworms learn a task. Please take a look at Frank, Stein & Rosen (1970) 169, 339-402. It is a fabulous debunking of 'chemical memory'.
"village of Garston" - unbelievable! I grew up in Watford. Garston isn't a village, its the bit of north Watford that is near(wish) junction 6 of the M1. It is horrible. I'm not surprised that a nerd from Garston is also a psychopath who enjoys killing small animals. Probably more fun than any entertainment Garston offers. (ps I haven't lived in Watford for decades but I know it very well)
The candela is a measure of human perceived brightness. As people's perception of the brightness of a light source depends on the wavelength of the light (we are most sensitive to wavelengths corresponding to green (around 550nm) and less efficient to shorter and longer wavelength) the definition of the candela has to include a multiplier representing the relative luminous efficiency as a function of wavelength - this function is known as V-lambda. The definition of V-lambda is based on people subjectively matching the brightness of lights with different wavelengths. The standard V-lambda defined by the CIE (Commission Internationale de l'Éclairage) is, in fact, based on measurements from a very small number of observers. There are different versions of V-lambda for daylight adapted (photoptic) vision, dark adapted (scotopic) vision, and the intermediate state - mesopic vision. The candela is defined in terms of the physical power of a light over an illuminated area at a single standard wavelength (and so is entirely physical at that wavelength), but to use the candela as a measurement of luminous intensity at any other wavelength one has to use a subjectively defined multiplier from the appropriate V-lambda. All pretty weirdly subjective for an SI base-unit!
(I study human vision for a living.)
The Register - Independent news and views for the tech community. Part of Situation Publishing
Join our daily or weekly newsletters, subscribe to a specific section or set News alerts
SUBSCRIBE
Biting the hand that feeds IT © 1998–2020
