This catch and release pattern is quite common in many wildlife surveys.
For example, when estimating tiger populations, instead of painting a white dot, rangers set up camera traps to take photos. Then they can use stripe patterns as a signature for subsequent re-appearances. Quite an interesting intersection for image AI with statistical counting methods.
Are there adjustments that need to be made for non-random sampling in that case? I'd imagine that with any territorial animal, a stationary camera is most likely to see the same animal multiple times.
The current government came into power riding on a wave of popularity, largely driven by their "IT cells". Now comfortably in power, they are probably dismantling the path, to avoid others who might try to follow.
The government is not just in a stand-off with big-tech companies. It continues to view its own citizens as adversaries with frequent internet shutdowns [1].
Can we stop with all this IT cell bullshit? Modi got in because he was more competent than the rajputra of the Nehru dynasty. My family are from Gujarat. We’ve known about Modi since he was governor. We moved out of our ancestral village several generations ago but still return on special occasions because our kuladevi is there. The area is semi-desert. Modi got water from the Narmada dam piped in. Now cash crops can be irrigated and the kids don’t die of dysentery any more. We used to have to stop at a larger town and walk a couple of miles down a dirt track to get there. Modi got a road built and regular bus service to the nearest city. That’s why he is popular. That area was staunchly Congress since 1947 and still is but you won’t find anyone with a bad word to say about Narendra Modi. He got things done and that’s what counts. The Hindu nationalism is just icing on the cake. (Remember there was a time when Congress also used to brag about how e.g Sardar Patel rebuilt Somanath.)
Now with the latest missteps such as demonetization and the COVID response etc. the bloom is off the rose a little but he is still genuinely popular and more importantly there isn’t anyone on the national scene “who might be trying to follow.”
He also led pogroms, segregated living areas by religion and promoted rioters and terrorists. Gujarats gdp per capita is only slightly above the median.
And IT cell is a reality that has been exposed several times over. Identical copy paste tweets and interviews with ex-it cell employees.
However, he didn't win elections because of them. His win in 2014 was a reaction to a government lead by the Congress, where the Congress party leaders were intent on defaming and destroying their own Prime Minister.
And the 2019 success was a result of the focus on base infrastructure, economic, diplomatic and defense policy in the first term (the government made genuine and very noticeable improvements in terms of sanitation, electrification and water access across the country).
They also made some epic economic blunders, however, that pain is truly being felt several years later (besides the immediate pain which they were able to explain away by promising long term benefits, but as those benefits don't materialize, they are losing trust), and appear to have given up on economic and social development, and appear to be focused on purely extending their rule through policies of hatred.
It remains to be seen how successful their strategy will be, but the initial results seem to indicate that they are greatly underperforming local elections. And most of these were conducted before the impact of teh devastating 2nd COVID wave, which the government didn't just mishandle, but were probably responsible for, became apparent.
Speaking of copy paste, what happened of the toolkit row?
Whether the objective was good or bad, should the people actively trying to trend things be called "IT cell for XYZ"?
It feels a bit hypocritical
I gave you actual examples of improvements I’ve seen with my own two eyes. People can’t drink gdp.
I don’t doubt there could be an IT cell. What I was reacting to was the idea that it is a major reason for Modi’s success. Like “Russian collusion” this is just cope by people who can’t accept that it is they who are the unpopular ones.
People don’t understand that propaganda like hypnotism only works on people who are already primed to believe it. Why does anyone listen to this BJP IT cell? Do you know?
I actually know Muslim friends in ahmedabad that cannot buy or rent a house in normal housing societies and have been ghettoized into Muslim enclaves with non existent municipal services.
I am glad that I live in the United states where nobody asks me for my religion or caste before renting or selling a home. United states is never going to be "dharmic", it got rid of its "dharmic" tendencies with the civil rights movement.
Edit: The dharmic comment was in response to another poster talking about United states becoming "dharmic".
The village I’m referring to (in Rajkot district) is less than 5% Muslim and those are Ismaili, Vohra etc. Even for them what happens to Sunnis in Ahmedabad might as well be on Mars as far as they care. What the RSS think about Hinduism is also irrelevant. Modi is revered for one reason and one reason only; what he actually did to make those peoples lives better. I mentioned in my original post that his popularity has not translated into popularity for the BJP in the area. That too is for local reasons.
LOL my “screed reads like an apologia for Hitler.” You are getting too emotional. I’m explaining why Modi is popular. Don’t think about it if it upsets you but don’t shoot the messenger in any case.
I suggest learning more about what the word dharmic actually means. It seems clear to me that you are using some non-standard interpretation of its meaning.
Because that’s where his political enemies are? Most politicians don’t have very deep reasons for why they act.
The set of people who worry about the state of the internet does not overlap very much with the set of people who worry if they have water and the set of people who worry about Hinduism is separate yet. In a democracy whoever can cobble together enough of these different groups (and many more) to make majority will win. If he cannot he will lose. It’s as simple as that.
Modi government is competent at marketing everything as his contribution, he is highly egoistic, doesn't accept any mistakes, or tries to push the blame to the Nehru Dynasty, Opposition etc. BJP IT Cell is one the most active proponent in pushing the image of Narendra Modi as the savior, whereas in reality, literally any leader who would have listened to experts instead of the election gurus, would have taken more sensible steps.
Modi government is full of yes man, and due to his popularity, the BJP party has itself become smaller than Modi, as elections are won in the name of the Prime Minister instead of the actual leader.
The government is also known for buying elected people post elections using money or power(CBI,ED intimidation) and known for manipulative numbers (they even changed the definition of GDP to make the numbers more attractive).
Modi actually follows Trump modulus of operandi, where instead of calling his opposition/critical news reports as fake news, he more conveniently calls them anti-national or pro-pakistani or western influences.
Anyways this stupid notion of no national alternative was also propogated the BJP IT Cell and BJP owned Media Houses only. Sadly a lot of Indians, including the commenter above have actually bought into it. If you actually look at numbers, India was falling behind on almost everything(jobs, economy etc) even before covid started.
Factually speaking, the worst term of UPA(2009-2014) was actually better than the modi government's first term in lot of matrices. Here is a one such study, you can find several others by googling - (please be aware of pro-BJP sources such as https://www.opindia.com, Republic TV etc)
https://www.livemint.com/politics/policy/manmohan-singh-vs-n...
he is popular because of the propaganda, by that logic hitler was also very popular. You are cherry picking data to suit your needs. Two examples at personal level do not balance the screw ups he has done at the national level. You are speaking as if no development work was done during earlier regimes. All the development started only after 2014. Stats speak another story about the development but then that is not what you want to hear.
Since you are still defending him, you must be blind to all other things seeing them as international conspiracy to defame your great leader.
> there isn’t anyone on the national scene
At this stage anyone, literally anyone, would be better than him
Uh Hitler was very popular. (To be precise he was more popular than the other guys.). The problem people have with Hitler isn’t that he was unpopular.
Really the amount of projection in some of these replies is incredible. All I’ve done is explain why one group of people have warm feelings for Modi (and FYI no development work was done in that area by prior Congress or BJP governments.) I’ve even acknowledged that he has made serious missteps. But as in parts of the US, in parts of India ORANGE MAN BAD is the only acceptable response it seems. Just a different shade of orange that’s all ;-)
> At this stage anyone, literally anyone, would be better than him.
If you say so. In the real world there actually has to be a person. That person has to actually have a platform, preferably something more than HES HITLER YOU GUYZ, and has to appeal to a vast an diverse electorate who will be driven primarily by parochial concerns. That’s how democracy works.
I understand that you do mention the pre-training / transfer learning approach clearly, but isn't it disingenuous to claim that you provide better performance based on (only) 100 labeled examples, when the pre-training dataset (Wikitext-103) actually contains 103M words?
Of course not. The use of pre-training on a large unlabeled corpus and subsequent fine-tuning is what the paper is about. It is stated repeatedly in the paper and the post.
It is totally correct and in no way misleading to say we need only 100 labeled examples. Anyone can get similar results on their own datasets without even needing to train their own wikitext model, since we've made the pre-trained model available.
(BTW, I see you work at a company that sells something that claims to "categorize SKUs to a standard taxonomy using neural networks." This seems like something you maybe could have mentioned.)
Got it. I was looking for input on how generalizable (the ability of weights to change/adapt) when the training labeled data is 100x smaller than the initial pre-training dataset?
Also, I don't understand the need to be so defensive though and the relevance between my employer and my post?
When you use the word disingenuous, you invited the response you got. Totally uncalled for to write that.
His response on your employer was likely driven by an assumption that you viewed this as free, open source competition to your product, and thus the negative comment.
To the OP:.
I've find a lot of NLP, and this is phenomenal work.
This was exactly what I was thinking, why complicate things by replacing the root volume when one can simply mount the disk to any other directory and point the application there?
Until this interesting video [1] by Tom Scott, I had never considered sand as a non-renewable resource. But now, it is clear that we now have one more resource that we might run out of.
In the context of today, yeah, there are tons of resources that we're quickly plundering our way through. The stories mentioned by Tom Scott and the parent article are tragic.
But, big picture, nothing on this planet is "non-renewable". Sand, for example, can be manufactured. As can petroleum, coal, clean water, etc. In fact all those products can be produced from the waste of the materials they are replacing. We can chuck a bunch of plastics into a pressure chamber and, with enough energy, regenerate the original petroleum products.
And that's the real crux: energy. We don't make our own sand and fossil fuels because it would cost too much in energy. But if we had more abundant sources of energy, the cost would go down, and more manufactured materials would become viable.
Our crisis is not that we'll run out of resources. We have plenty. Our predicament is a lack of energy sources. We make up for it by consuming energy our planet has stored, in the form of fossil fuels and other so-called "non-renewables". The solution is to transition ourselves away from this handicap. We need to have more sustainable and abundant sources of energy. That pretty much amounts to solar, as there are no other truly sustainable sources of energy.
So the real solution has not much to do with using less "non-renewables" and more to do with improving our abundance of solar energy.
Yes. I suppose my comment kind of comes off as saying something obvious. Obviously if we had infinite energy we wouldn't have any problems.
But really what I was trying to get at was that our focus, with respect to the environment, is perhaps in the wrong place. We focus so much on reducing use. Really our focus needs to be on how we can get _more_ energy.
It's like in a business. Don't focus on the cost centers, focus on the profit centers. How can we drive _more_ business rather than making the existing business more efficient. For a growing company that's usually the best advice.
So why, as a growing species, are we optimizing our energy usage? Screw that. Let's get more energy! Let's blanket the land with solar panels so we have enough "fuck off" energy to do whatever the hell we want. In particular, enough energy to get into space and build a Dyson sphere so we can get even more energy.
Imagine if we took all the money and time invested into optimizing energy usage, and instead had spent it on solar panels?
It's all really counter to public opinion. Environmentally conscious people love their LED lights. I do too. But I also love optimizing systems, and optimizing systems is often not what's smart for a business.
> But really what I was trying to get at was that our focus, with respect to the environment, is perhaps in the wrong place. We focus so much on reducing use. Really our focus needs to be on how we can get _more_ energy.
The article mentions, among other things, thieves who steal sand, sell it, and bribe police to look the other way. While that's the state of affairs, lowering the cost of energy is beside the point.
No, it would be profitable for them regardless. You might live in some science fiction world where energy and automation were so abundant that manufacturing sand from something other material and shipping it in made economic sense.
cost = energy-intensive manufacturing + lots of shipping
It wouldn't prevent this crime because it's never going to be as cheap as going to the river and filling up a truck.
cost = a short drive + a few bribes
Technology is absolutely not the solution to everything (although sometimes it helps a lot).
How do you propose sourcing Helium? Barring a few atoms here and there from fusion experiments and H-bombs I don't know of a good source without leaving the planet.
If the natural production of Helium via radioactive decay in the Earth isn't enough, we can use create it from Boron and Lithium after using copious amounts of energy to accelerate protons into them, or for just Lithium, Deuterium.
A similar eye-opening for me was when I've learnt that Saudi Arabia is importing sand and camels from Australia. I couldn't find a reliable enough source now with a quick search, but here's a similar story on BBC: "Even desert city Dubai improts its sand. This is why." http://www.bbc.com/capital/story/20160502-even-desert-city-d...
I am from Kerala, in India. The amount of damage the illegal sand-mining done here in rivers is pretty alarming.
I think we should adopt living in smaller houses, which are built out of locally available materials, instead of cement concrete. As far as I know, concrete buildings do not have a lifetime exceeding 40 years, where as older construction techniques do give us longer lasting buildings.
Local concrete used in India include a mixture of slaked lime, honey, water chestnuts, rice husk etc. - e.g. [1] Seems to last a couple of centuries, at least.
Interesting - comments elsewhere here are disparaging about wood and laminates due to perceptions about longevity, while the stuff concrete is replacing in India was hardly short lived.
Seems like the #1 reason we are running out of sand, is concrete doesn't last very long. The #1 reason for concrete not lasting long that is rebar. Sure rebar lets you use less concrete, but then it doesn't last nearly as long. Without rebar concrete can last 1000s of years (like the roman aqueducts.)
I'm fairly certain Roman concrete also relied on chemistry as an alternative to thermal energy in its manufacturing process, as thermal energy simply wasn't as abundant to the Romans.
The longevity may simply be a fortuitous consequence, though a highly significant one.
Roman engineering was also rather more overengineered. You compare their columns to modern or say gothic ones, and they clearly have a lot of excess strength.
What I understand is that desert and beach sand tends to be old/heavily weathered sand that is too fine with rounded edges. And is physically weak. Sea sand also tends to have chlorides which causes problems.
The foundation of the house I'm living in was probably made with beach sand (1908 construction). One wall that gets wet frequently is turning to dust.
Flip-side of that is the timbers are all old growth redwood. If you ever poke at the stuff after 100 years as long as it didn't get wet it's fine. 50-50 chance whoever buys this place will replace the foundation. And hopefully with proper concrete. Do that and the building will last into the next century.
The point is supposedly that providing housing for the current exploding population is more important/urgent than their great-grandchildren having to tear down and rebuild their houses in 100 years.
There is a beach in south Sweden which was mined for sand, I think it was for biosand filters for water purification. The said they exported sand to Saudi from there. I presume different sands have different properties.
For example, when estimating tiger populations, instead of painting a white dot, rangers set up camera traps to take photos. Then they can use stripe patterns as a signature for subsequent re-appearances. Quite an interesting intersection for image AI with statistical counting methods.