It's 15 companies that adopted a technique and it's part of a broader constellation of experiments in this area which have been confounding the traditional logic that productivity would decrease. https://www.4dayweek.com/research
The people representing the businesses in the preprint hold titles like co-founder, CEO/founder, COO, General Manager, and CEO. The size of the business and sector are also noted. I think your framing of them as "unidentified people" is therefore off, it is certainly not the same as a journalist conveniently using "unnamed sources", this is standard academic practice.
Different companies measure different things, but they do measure and that is addressed in the paper. "revenue (DM10), profit (DM4), other financial targets (DM2, DM6), customer/client satisfaction ( DM8, DM6), story points (DM14), sprint goals (DM7), billable hours (DM12), capacity ( DM4), response rates (DM10), standard operating procedure metrics (DM9), sick leave (DM1, D M 4. DM9, DM15), lodgements (DM12), employee happiness (DM6, DM15), projects delivered on time (DM15), and net promoter score ( DM4)". There were also other benefits like hiring and retention.
So this is not what "unidentified people" "thought" about productivity, this was founders and the c-suite using their existing favoured metrics. On those metrics a large number of them reported an increase in productivity, and a larger group reported no deleterious effects on productivity. This is broadly consistent with the trends in the wider research into this area globally, which continually go against the predictions that productivity will drop. Is it universally applicable? I don't think anyone is claiming that.
I've followed this area for a while and, sorry to be impolite, it is your summary that is less accurate than the the one you accuse of being AI-generated.
Having read all of the above, I think it is fair to say that this is a study about what these people "thought." Just because their thought involves some homegrown, personally-favored metrics, doesn't change the fact that this is a qualitative survey report.
Feta comes from φέτα/~"slice", so it's not named after a region. Melbourne has one of the largest Greek populations in the world. Why don't these Greek people have right to their heritage? If the British were to start trying to say that "sandwich" was protected and American sandwiches were "inauthentic and deceptive" would you take such a claim seriously?
> Why don't these Greek people have right to their heritage?
Probably because it would be difficult to find any reasonable middle ground between “can only be made at the geographical origin” and “can be made anywhere by anyone”
Also for a lot of product the origin more than the heritage of the people is central, such as the climate and soil in Champagne (which perhaps soon will be most historically authentic in southern Sweden after some climate change).
This is about regions keeping the right to their products, not necessarily people retaining that right. Move from Champagne and you can’t make Champagne. Not that complicated. Feta is a regional produce too - the name doesn’t really change that.
> If the British were to start trying to say that "sandwich" was protected and American sandwiches were "inauthentic and deceptive" would you take such a claim seriously?
But the poster you challenged was specifically complaining about a generic product name that was not a place name. Because it opens Pandora's box in terms of every generic food name being reclaimed by the place it originated.
The list of place names which are also products, and the rhetorical ease of defending their protection for such cases, does not make the argument about protecting local generic names as well, precisely because it is not as easy to defend such names. What criterion would you use? The degree of feel-good small-town credentials of the claimant?
I agree there may be a subtle difference between “actually geographic” and “traditionally regional but generic” but I don’t think it’s actually important. That Pandora’s box seems well worth opening.
If a product was made exclusively in a region for some amount of time (say a few hundred years, and nowhere else) then I think that’s a pretty strong case for protecting that tradition in the region whether the produce bears that name or not.
Possibly, but “dishes” and “exportable products” seem a bit different from an industrial perspective. I don’t think dishes will ever be up for discussion in this context.
"Feta" is not the heritage of Greek people. It's a Protected Designation of Origin that covers specific geographical locations in Greece, the regions of Thessaly, Thrace, Epirus, Macedonia, Central Greece, Peloponse and Lesvos. Greeks, living in Greece, outside of these regsions, cannot sell their white sheep's milk cheese as feta. For example Cretan cheesemakers, 100% Greek themselves, can't sell their white sheep's milk as feta and must sell it as "white cheese" instead.
Now, if Greeks, living in Greece, making cheese with the milk of Greek animals can't call their cheese "feta" why should Australians whose grandparents came from Greece be able to?
Censorship and demonetisation on YouTube seems to be happening to people fairly regularly, and without appeal, via poorly calibrated algorithms, corporate pressure and to satisfy mobs of complaining people for various reasons. The rules seem to be vague to the point of meaninglessness and it seems even the stars of the platform who often share managers and production companies with other stars have difficulty getting in touch with YouTube to resolve issues.
I think most content producers have wised up to this and diversify their audience over multiple platforms. If you rely on the income putting all your eggs in the YouTube basket is a massive risk until they clean up their processes (there are many automated and social methods to do some of this but they seem completely uninterested in doing it).
> Censorship and demonetisation on YouTube seems to be happening to people fairly regularly, and without appeal, via poorly calibrated algorithms
Demonetization happens when a video doesn't meet their "advertiser-friendly" policy, but there is an appeal process where you can have a human look at it to determine if the original assessment was wrong [0]. Do you have data to support your claim that their algorithms are "poorly calibrated"?
Can you provide some examples of "censorship"? They do have policies that things like graphic content or spam is not permitted and will be removed from the site, but I think that's reasonable.
> It seems even the stars of the platform who often share managers and production companies with other stars have difficulty getting in touch with YouTube to resolve issues.
YouTube provides email support with a 1-business day response time to all creators [1], and the bigger channels get their own Partner Managers [2].
When Pewdiepie (person on your [2] link on Silver & Up banner) complains [0] that YouTube isn't communicating enough, most people will agree that problem is at their side.
>Demonetization happens when a video doesn't meet their "advertiser-friendly" policy, but there is an appeal process where you can have a human look at it to determine if the original assessment was wrong [0]. Do you have data to support your claim that their algorithms are "poorly calibrated"?
It happens all the time when content creators are using content under fair use. Jim Sterling goes over demonetisation with ContentID [1] and latest changes [2] in his videos. I don't know if he have used that email support, but it would be fair to assume that he have tried and gave up.
> YouTube provides email support with a 1-business day response time to all creators [1], and the bigger channels get their own Partner Managers [2].
Multiple demonetized channels have stated that they have not received responses via the official support channels. It's all well and good stating a 1-business day response time, but if Youtube doesn't follow through that, where does that leave the content creators?
It's like businesses with support response SLAs that are cleared by the sending of a robo-email from their support system. No actual support has been rendered.
https://www.maxlaumeister.com/blog/google-is-deleting-your-f... has some, but anyone who's been following it knows it's happening beyond that. I am surprised that infinitesoup is apparently unaware of these issues given his account on HN is 591 days old and over it's history has posted about absolutely nothing but YouTube. Indeed an account on Reddit called infinitesoup, possibly unrelated I concede, has wall-to-wall comments on YouTube too. They all read like a superfan or possibly an employee of Google providing support. The usual etiquette on HN is to simply disclose an interest and then argue a point. While it's not against the rules I don't like having to do my own research and finding out the person posting links from Google support is likely a Google employee (the person connected with the Reddit account seems to be a Google employee, maybe this guy isn't).
Even if they are different accounts and this guy isn't a Google employee they haven't presented any other credible evidence for this wonderful support anyway. They have only presented a policy aspiration. Google is not Amazon; Google is notorious for bad customer support. Our prior belief for "will YouTube provide good support" shouldn't be very high given they are part of Google. Therefore it doesn't take many data points in the direction of poor support to confirm that.
There are clear and sometimes good reasons for what Google does, it's a great company with competent employees but let's not drink the Kool-Aid and pretend they have great customer support just because there is a policy document aspiring to have good response times.
Play store has had the same issue for years, and there really isn't an alternative either. You can try to diversify, but nothing compares to the android crowd due to low, low cost devices. Google doesn't want to pay for decent support for something making them a ton of money, why would they do it for YouTube? You're stuck. Play the game and play it well.
What are we up to now? Three preloaded spyware scandals, possible remote execution via the Intel stack and now this vulnerability. That's just what we know about, who knows what else exists. I don't think I can buy another one, which is sad as I think it was a timeless and great design.
I plan on using my quad core T520 for probably another 5+ years. All of their laptops after the T520 series have the full size keyboard with numberpad which off-sets the center of the keyboard, so now your typing is mostly happing on the left side of the keyboard and that causes wrist strain.
Having a numberpad is really lame on a laptop. I won't buy one and I know of no one else that likes the numberpad either.. sadly many manufactures are doing the same.
Wow this was actually going to be a major buying factor in my next Laptop, was really considering a gaming laptop for the numpad but I guess it wouldn't be too difficult to buy a bluetooth numpad for the right side of the keyboard/get used to the numbers above the keyboard. Thanks for this angle. Never considered wrist strain.
Which would involve needing to reinstall the OS, or at least the kernel, because Trisquel strips out the kernel drivers needed to update the microcode.
I own a couple of Apple notebooks. Two years ago I bought a used X201 because I needed a Linux backup (official excuse) and because I love the design (actual reason). It's completely different from Apple's approach and Richard Sapper's original draft is still visible in those machines.
And it runs Ubuntu just fine with an SSD upgrade and still enough memory for all purposes I have.
I think in addition to histograms and features like SIFT, SURF, and others like DAISY[0] that would permit searching with images as a query it would be beneficial just to use a neural network to text index using the classes at very top of a neural network, although you are right some features in layers just below that could be used too as I understand it.
You could then use classical text indexing on the text, perhaps with a topic model like LDA. Then an image with a plane in it will be indexed by "plane" via the output of the neural network but would also come first, or in the top results, when using "flight" as the query via a topic model.
Ditto for word2vec or para2vec over those words, the benefit being you can bring the knowledge of relations contained in the textual training data, Wikipedia or something else, to bear on the problem. I.e. a golf club and a baseball glove might not be correlated in the neural network that annotated the images but might be correlated in the text based knowledge model trained on Wikipedia and so a query of "sport" might bring both images up.
The big players like Google already have caption generation that's capturing relationships between objects.[1]
Here are two Facebook posts from Taleb related to the topic and a ~3 minute video that has the main point.
"The establishment composed of journos, BS-Vending talking heads with well-formulated verbs, bureaucrato-cronies, lobbyists-in training, New Yorker-reading semi-intellectuals, image-conscious empty suits, Washington rent-seekers and other "well thinking" members of the vocal elites are not getting the point about what is happening and the sterility of their arguments. People are not voting for Trump (or Sanders). People are just voting, finally, to destroy the establishment." - 7th March - https://www.facebook.com/nntaleb/posts/10153654273663375
"What we are seeing worldwide, from India to the UK to the US, is the rebellion against the inner circle of no-skin-in-the-game policymaking "clerks" and journalists-insiders, that class of paternalistic semi-intellectual experts with some Ivy league, Oxford-Cambridge, or similar label-driven education who are telling the rest of us 1) what to do, 2) what to eat, 3) how to speak, 4) how to think... and 5) who to vote for.
With psychology papers replicating less than 40%, dietary advice reversing after 30y of fatphobia, macroeconomic analysis working worse than astrology, microeconomic papers wrong 40% of the time, the appointment of Bernanke who was less than clueless of the risks, and pharmaceutical trials replicating only 1/5th of the time, people are perfectly entitled to rely on their own ancestral instinct and listen to their grandmothers with a better track record than these policymaking goons.
Indeed one can see that these academico-bureaucrats wanting to run our lives aren't even rigorous, whether in medical statistics or policymaking. I have shown that most of what Cass-Sunstein-Richard Thaler types call "rational" or "irrational" comes from misunderstanding of probability theory." - 9th March - https://www.facebook.com/nntaleb/posts/10153658794008375
Significant perhaps in the statistical sense. The domain is the same. I believe he is also a professor of mathematics. He has some eminence in this area in any case.
The examples where this happens have always seemed fairly weak to me. How many of the grave errors, not just where it's the wrong type of animal or container but actually thinking it's radically different, survive an application of Gaussian blur? Furthermore self-driving cars are a combination of signals; you are going to need to simultaneously fool both LIDAR and cameras.
On top of that you are going need to fool them over multiple frames, while the sensors get a different angle on the subject as the car moves. For example in the first Deep Q-learning paper, "Playing Atari with Deep Reinforcement Learning"[0], they use four frames in sequence. That was at the end of 2013.
I don't think anyone will be able to come up with a serious example that fools multiple sensors over multiple frame as the sensors are moving. Even if they do then inducing an unnecessary emergency stopping situation is still not the same as getting the car to drive into a group of people. Even if fooled in some circumstances the cars will still be safer than most human drivers and still have a massive utilitarian moral case in relation to human deaths, on top of the economic case, to be used.
The fooling of networks is still an interesting thing, but it's been overplayed to my mind and is not particularly more interesting than someone being fooled for a split second into thinking a hat stand with a coat and hat on it is a person when they first see it out of the corner of their eye.
1. Gaussian blur is just a spatial convolution (recall from signal processing). If a network is susceptible to adversarial examples, it will still be susceptible after a Gaussian blur (assuming the adversary knows you're applying a Gaussian blur. If the adversary doesn't, that's just security by obscurity, and they'll find out eventually).
2. A sequence of frames does not solve the issue because you can have a sequence of adversarial examples (although it would certainly make the actual physical process of projecting onto the camera more difficult, but not really any more difficult than the original problem of projecting an image onto a camera).
3. Using something conventional like LIDAR as a backup is the right approach IMO, and I totally agree with you there. But Tesla and lots of other companies aren't doing that because it's too expensive.
1. If that's the case perhaps another kind of blurring? "Intriguing properties of neural networks" (https://arxiv.org/pdf/1312.6199.pdf page 6) has examples where you get radically different classifications that I don't think would occur naturally or survive a blur with some random element, let alone two moving cameras and a sequence of images. As the title says it's an intriguing property, not necessarily a huge problem.
2. I honestly can't think of a situation where this could occur. It's the equivalent of kids shining lasers into the eyes of airline pilots, but the kids need a PhD in deep learning and specialised equipment to be able to do it. A hacker doing some update to the software via a network sounds much more plausible than attacking the system through its vision while it's traveling.
3. This is the real point in the end I guess, this Google presentation (https://www.youtube.com/watch?v=tiwVMrTLUWg) shows that the first autonomous cars to be sold will be very sophisticated with multiple systems and a lot of traditional software engineering. Hopefully LIDAR costs will come down.
1. Those are examples for a network that does not use blurring. You have the be careful because, remember, the adversary can tailor their examples to whatever preprocessing you use. So the adversarial examples for a network with blurring would look completely different, but they would still exist. Randomness could just force the adversary to use a distribution over examples, and it could mean they are still able to fool you half the time instead of all the time. However, I wouldn't trust my intuition here: that is really a question for the machine learning theory researchers (whether there is some random scheme that is provably resilient or if they're all provably vulnerable, or proving some error bounds on resilience, etc.).
2. The problem of projecting an image onto a car's camera already implies you'd be able to do it for a few seconds.
"It unlikely to happen" is not a good strategy to rely on with systems operating at scale. There are about a billion cars on earth traveling trillions of miles every year, many of which will eventually be self-driving. At that scale, you don't need a malicious actor working to fool these systems, you just need to encounter the wrong environment. And even if the system is perfect on the day it's released, that doesn't mean that it will remain so indefinitely (even with proper maintenance).
Studying induced failure in neural networks may help us understand the failure modes and mechanisms of these systems.
The full results in the report seem to me be somewhat more complex to interpret than the summary, or the original submission title, suggest. On the first question (Q18jt), a 4 point scale without a "don't know" neutral option, there are large numbers of people putting global citizenship before national citizenship (51%-49%). However, when you get to a later question (M2), that asks people what their most important identity is, national citizenship overtakes global citizenship overwhelmingly (52% national, 17% global, 11% local, 9% religious and 8% race/culture). National citizenship is the largest in all countries besides three, and it's the second largest in two of those. When given a variety of choices in M2 the level of people supporting global citizenship (17%) as a primary identity is even lower than even those who "strongly agree" to global citizenship being more important to national citizenship in first question (22%).
In M2 it seems all of the "somewhat agree" and a few points of the "strongly agree" have gone elsewhere including back to national citizenship which creates a tension between the results of the two questions. It’s quite stark in the case of India where in the first question 67% place global citizenship over national citizenship but in M2 51% have national citizenship as their primary identity and only 6% have global citizenship.
I think the first question (Q18jt) would be more interesting if it was a 5 point scale with a "don't know" option provided rather than trying to railroad people into an answer with a 4 point scale. Even then I think the M2 question is a better instrument as it provides a clear set of choices. For example Pakistan goes from 56% global in the first question to 2% global in the second. Most of those seem to move to religion as their primary identity (43% answer religion on M2). Many religions are globally oriented toward all humanity but the M2 question lets us see more clearly what they believe.
Even though the second question (M2) is better it probably could still do with some examination. For example the only country where global citizenship was higher than national citizenship in M2 was Spain, but was that because many people in Catalonia and the Basque Country are choosing global over "Spanish" because the "local" option was translated in a particular way? If they used a word closer to "neighborhood" you could see how that might happen. Similarly was "local" translated more broadly in the Indonesian translation and became the dominant answer there (56%) as people could identify with an island/state/ethnic/linguistic group that was "local", like Bali for example. It's hard to tell without all the data and the exact translations of the questions.
The people representing the businesses in the preprint hold titles like co-founder, CEO/founder, COO, General Manager, and CEO. The size of the business and sector are also noted. I think your framing of them as "unidentified people" is therefore off, it is certainly not the same as a journalist conveniently using "unnamed sources", this is standard academic practice.
Different companies measure different things, but they do measure and that is addressed in the paper. "revenue (DM10), profit (DM4), other financial targets (DM2, DM6), customer/client satisfaction ( DM8, DM6), story points (DM14), sprint goals (DM7), billable hours (DM12), capacity ( DM4), response rates (DM10), standard operating procedure metrics (DM9), sick leave (DM1, D M 4. DM9, DM15), lodgements (DM12), employee happiness (DM6, DM15), projects delivered on time (DM15), and net promoter score ( DM4)". There were also other benefits like hiring and retention.
So this is not what "unidentified people" "thought" about productivity, this was founders and the c-suite using their existing favoured metrics. On those metrics a large number of them reported an increase in productivity, and a larger group reported no deleterious effects on productivity. This is broadly consistent with the trends in the wider research into this area globally, which continually go against the predictions that productivity will drop. Is it universally applicable? I don't think anyone is claiming that.
I've followed this area for a while and, sorry to be impolite, it is your summary that is less accurate than the the one you accuse of being AI-generated.