GrantS

I would argue the "significant landmark" problem has essentially been solved as a side effect of online photo sharing sites (originally Flickr, now everything) -- the most frequently photographed things are the landmarks, and the number of photos scales with the significance of the landmark. When you search for "Rome" on Flickr, the clusters of photos that pop out as being 3D-reconstructable are precisely the landmarks.

See this popular work from 2009, "Building Rome in a Day": http://grail.cs.washington.edu/rome/

Quoting: "The data set consists of 150,000 images from Flickr.com associated with the tags "Rome" or "Roma". Matching and reconstruction took a total of 21 hours on a cluster with 496 compute cores. Upon matching, the images organized themselves into a number of groups corresponding to the major landmarks in the city of Rome. Amongst these clusters can be found the Colosseum, St. Peter's Basilica, Trevi Fountain and the Pantheon."

If anyone wants to visualize how a variety of basis transforms (including Fourier/DFT, Eigenvectors/PCA, Haar Wavelets, Random basis, and Identity basis) can be applied to images, I created this a few years ago which seems to have become the #2 google result for "basis transform":

http://www.cc.gatech.edu/~phlosoft/transforms/

It's a bit dated by being a Java applet, but source code is all there, written in Processing. (I should probably update it now that processing.js is a thing. Years pass quickly.)

Here are the STL and Blender files for a different 3D printed clock designed by Christoph Laimer in case you'd like to print your own:

http://www.thingiverse.com/thing:328569/#files

This 5-minute video is really worth watching as he explains how it was designed, how it functions, shows a time-lapse of it being assembled, etc.:

http://www.youtube.com/watch?v=HgZBPYJ2Y-w

To each his own, but I just want to point out that your examples of web searches and email require that SOME third party be involved because they are, by definition, attempts to reach a computer somewhere else on the internet.

In contrast, searching the local files on one's own computer or launching applications on one's own computer are private, local events that need not be shared with any third party.

Inherently private stuff just happens to leak out into the world now because Spotlight does web search too.

One thing I love about reading old papers/essays/articles is that it makes it so clear the particular things we take for granted now. It is remarkably hard to step back and get that objective point of view about the current state of technology with no help from the past.

> They will remember that such-and-such a person did some possibly relevant work on a topic of interest back in 1947, or at any rate shortly after World War II, and they will have an idea in what journals it might have been published.

For all that the title evokes visions of cyborgs, Licklider here is really just asking for a search engine and the ability to plot some points on a graph, which is beautiful because these are things I do everyday, and they are indeed so useful and commonplace that I take them for granted, but even I couldn't always do these things in my lifetime.

I also found the following incredible as a train of thought:

> Yet there is continuing interest in the idea of talking with computing machines. In large part, the interest stems from realization that one can hardly take a military commander or a corporation president away from his work to teach him to type.

Very cool! Some hopefully helpful feedback: Running Safari on a Mac, and the text cursor visually shows up in a different place than the actual editing occurs, making editing tough. It is off by an amount that varies largely when I change the size of the on-screen text, sometimes off by 1 character, sometimes off by 4 characters. At a really large font size, I got it to mostly align and it made editing possible.

PBS aired a fascinating documentary earlier this year about an orangutan that was taught sign language in the 70s by being raised on a college campus as a "human child", explores both the personal side and ethical issues. I was surprised I had never heard of him, and even more surprised to find that he now lives at the zoo a few miles from my house. I had no idea.

Video here, I'm not sure if this plays outside the US: http://video.pbs.org/video/2365286726/

I almost didn't read the article due to the headline but the fMRI studies (which are really the focus of the article) are fascinating, and there is a surprisingly deep discussion about the line between idioms and metaphors -- particularly how this affects whether the brain engages the motor cortex when processing language and how it might vary by individual.

Note that the headline isn't actually implying machines can't be intelligent but that their internal states won't correspond to human states unless their cognition is grounded in a similar set of sensations. But this is already the case with humans from vastly different cultural contexts. The words coming out of your mouth only mean the same thing to others to the extent that they share a similar history of experiences when interacting with the world.

So I can certainly believe that an AI whose internal concepts are based on embodied/simulated experience would seem more relatable than one raised purely on books, but that's true of humans too so no big surprise and not an insurmountable barrier as one of the quoted sources in the article suggests. Non-embodied agents will speak in idioms and embodied agents will speak in metaphors.

Thanks for the link, but perhaps you meant last month? The tear down appears to be from 33 days ago.

Indeed, architects have definitely been using the Unreal Engine for quite awhile, even before it started looking this good. Here's an article from 2007:

http://www.businessweek.com/stories/2007-12-21/unreal-archit...

It is never made explicit in the video but imagine that corporations are the current dominant super-organism on which the economy depends. Humans are only designing and building cars because corporations have no better way of getting that job done. Thus, humans are used by corporations just as horses were used by humans before the corporation became the dominant economic species on the planet. Corporations may not always need humans to survive.

Just as cars replaced horses for the benefit of human masters, robots may replace humans for the benefit of corporate masters.

You may still disagree with the analogy but I think this is at least the proper analogy to be arguing about.

This work comes out of Harvard's Wyss Institute for Biologically Inspired Engineering, which was not on my radar before today. Apparently just last week they released some even more impressive work (in the sense that I would not have thought it possible) on a self-folding mobile robot: http://wyss.harvard.edu/viewpressrelease/162/robot-folds-its...

The video of the robot self-folding (and walking away!) at about 1 minute in blew me away, highly recommended to anyone who missed this story last week.

Ha, yes, that is a pretty accurate assessment of what they did there.

It's more of a motivation or justification of the work than anything -- they want to apply the paper's methods to control marine-related engineering systems which are becoming more important to the future of Taiwan.

It sounds like a non-sequitur because it was probably copied and pasted from the grant proposal to get funding for the work. The acknowledgments show that funding came from the "National Science Council of the Republic of China, Taiwan", so that organization would naturally want to know why they're funding work on fuzzy numbers and how it benefits the country.

Strangely enough, I just started getting spam from SAGE this morning about subscribing to their Sociology journals. My area is computer science so it seems they're just spamming university email addresses. Not too classy.

Using the lower level AVFoundation classes to start a capture session, any app could absolutely get live camera access to either front or back camera without the user knowing anything. I know because my app shows a live preview to the user right when the app starts up, no permission required, direct access to pixels, and if I chose not to display the UIView that shows the live camera data, that would not change my app's access to the data (under iOS 7 and earlier, I mean).

It's possible Apple screened for non-camera apps using those APIs to keep spying apps out but (Edit: bad example, this was an Android app. I had said: "there was that flashlight app that was storing/reporting user GPS locations without permission so obviously things were slipping through the old system.")

You can, of course, navigate to settings and toggle the switch for anything you accidentally denied, but I wonder how many users know that. Also, unless a developer explicitly adds some message reminding the user that they have already denied access to some resource, the user may very well forget that they ever denied anything and think that the app just doesn't work.

All that said, I think Apple is making the right choice by favoring privacy and security even if it's at the expense of a completely frictionless user experience.

I had somehow missed the news from WWDC that accessing the camera now requires explicit permission from the user.

I'm really surprised camera privacy took this long to arrive, and I wonder what the reasoning was for not implementing it at the same time as microphone permissions arrived in iOS 7.

One of the most incredible parts is that they've already run feature detection on all 100M images/videos and extracted 50TB of:

"SIFT, GIST, Auto Color Correlogram, Gabor Features, CEDD, Color Layout, Edge Histogram, FCTH, Fuzzy Opponent Histogram, Joint Histogram, Kaldi Features, MFCC, SACC_Pitch, and Tonality"

The good part about this for researchers is not only that this saves dozens of CPU-years of computation (back of the envelope, it would take 15 years for my laptop to extract those SIFT features alone), but that any differences in learning/recognition performance on the dataset can be attributed to the algorithms in question, uncomplicated by which researcher engineered the best features for the dataset. On the other hand, it's a challenging dataset to work with because you can't just download it and process it locally as has been traditionally done. I'll be interested to see how many take advantage of it.

This seems logical, especially in light of the Aperture news, but I'm not so sure -- all references to the new Photos app, including the demo at WWDC, refer to it as a viewer for your iCloud-stored photos. It's possible that instead of cloud storage being an option added to iPhoto, it is being re-designed as the primary use case, with your Mac as just another iPad-like device that may occasionally cache some local copies of photos as needed. I understand that from a practical point of view in 2014, but I think I'm philosophically against it.

Watching the video, it was odd that he had absolutely nothing to say about potential uses of AI when asked that question. I'm assuming it was either a question so out of left-field that he just wasn't prepared or interested in discussing it and distracting from Tesla/SpaceX, or that he has lots of ideas he doesn't want to make public yet.

More understandable that he repeatedly says he doesn't know how to stop unfriendly AI, but there is always this: http://en.wikipedia.org/wiki/Friendly_artificial_intelligenc...

Right, while disruption is theoretically good for consumers, the major effect I've seen from Amazon over the past 15 years is that nearly every physical bookstore around me has closed. I would prefer a world without Amazon to a world without bookstores.

My city is getting more of these and I always check before turning right. But last week I witnessed a car-bike collision that I'm unsure how to ever avoid: traffic in my lane was backed up to a standstill, but oncoming cars were flowing freely and one took a left into a restaurant parking entrance (between two stopped cars in my lane) just as a bike was passing the stopped cars in my lane on the right (NOT in a designated bike lane but even if it was, the left-turning car would have had no way of seeing the bike coming because all cars were stopped and blocking view.) Bike hit side of turning car at full speed. Neither saw each other coming. Bicyclist was seemingly ok and kept riding after 30 seconds. Do you just outlaw left turns in the presence of bike lanes?

As I said in this case there was no bike lane and my understanding is that this is a known danger and the cyclist should not have been passing all the stopped cars on the right. See example #8 on this page: http://bicyclesafe.com/

Interesting! B&W makes total sense. Low res also makes sense for head tracking but I guess that means pupil tracking is out of the question. Regarding low framerate, like lower than 30 fps? I would have thought continuous perspective shifting would require a framerate at or better than screen refresh rate to appear normal to the user.

As a computer vision researcher, I am impressed that Amazon has made the most computer-vision-centric mobile device seen to date. I have no idea if it will catch on, but having a dedicated physical button for object recognition is huge insofar as it means they actually intend for people to be using this all the time, so they'll only be devoting more and more energy to recognizing everything in the world. Sure, it's mostly so you can then buy those things from Amazon, but I have to think there will be other useful applications given the provided recognition SDK.

I'm also curious if developers will have direct access to the four corner cameras on the front, and if the head-tracking it uses for 3D visual effects includes detailed eye/pupil-tracking -- if it doesn't yet, you can be sure that's planned for the future.

Edit: So the dedicated recognition button will already pull up wikipedia entries for recognized works of art. I wonder if they'll ever let third party developers hook into the hardware button so that third party plant-recognition, insect-recognition, action-figure recognition (etc) app results can be called up with that single button push in case Amazon doesn't recognize a purchasable item.

From the Kickstarter description, they're leveraging existing open source software: "The onboard software is based on the 3D Robotics open-source code"

Also note, they DO NOT include hazard avoidance in the current product: "The reason for this is that there is no avoidance system included in the first version."

So I think your questions are legitimate. The underlying question is: do we now live in an era when a team of mostly non-technical people with a couple of smart hackers can put this kind of product together from existing technologies without inventing anything really new? I guess the answer is mostly yes but I wonder if obstacle avoidance is really something that can be left out without serious consequences.

Take a look at Francis Bacon's Novum Organum written in 1620, specifically the sections on the four idols of the human mind that sometimes prevent one from reaching truthful conclusions. To my mind, this is philosophy of science before any practices had really crystallized. Reading this in college, I remember being impressed that Bacon seemed so modern in his thinking 400 years ago. His awareness of and focus on the fact that cognition is influenced by cultural context is a huge step and I was unaware it had been made so long ago.

http://en.wikipedia.org/wiki/Novum_Organum

Full text here: http://www.constitution.org/bacon/nov_org.htm

"It was intended to compete against or replace Google Maps for iPhone users, but it turned out to be filled with errors and distortions, some of them hilarious — the Brooklyn Bridge imagery was so twisted it looked like it was collapsing into the East River."

It bothers me that the genuine usability issues stemming from some incorrect landmark data at the launch of Apple maps are still confused with eye candy hiccups due to the fact that pixel-perfect automated city reconstruction is a hard computer vision problem. The distortion of the Brooklyn Bridge's imagery when projected onto noisy height-map data had no effect on routing people to their destinations correctly. You'll find countless examples of funny-looking projections, weird shadows, etc. on Google Earth as well simply due to the nature of projecting imagery captured from one viewpoint onto incomplete proxy geometry and viewing from another viewpoint.

Mean shift clustering, for one. This short PDF provides better insight than wikipedia. Does NOT require you to specify the number of clusters like in K-means:

http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/TUZE...

http://en.wikipedia.org/wiki/Mean-shift

Example python code using scikit-learn: http://scikit-learn.org/stable/auto_examples/cluster/plot_me...

I'm curious what particular claim you think Hofstadter is pushing that is unwarranted -- I don't see how a book of wandering musings could earn the label "pseudoscience" when it doesn't claim to be scientific in the first place (as far as I recall -- it's been about 15 years since I read it). It's a bunch of interesting ideas he ties together in clever ways.

The best I can come up with is that you're talking about his idea that self-reference lies at the heart of consciousness, which I think an intelligent reader can see is just a guess of his. Whether or not it's true, we can still see how it's useful to think of ant colonies as organisms, and we still learn about the incompleteness theorem, and we now understand what a fugue is and how it has textual and visual analogues, etc.

Maybe there's a group of people who treat GEB's claims differently than I do that I'm unaware of?