axelroze

You can limit the search query per website in DDG (and probably in others)

Example: `rust slow compilation site:stackoverflow.com`

Wouldn't the dethroner of Google be some new technology which is not a search engine like Google but better at solving the original task of finding information on how to solve problems?

Just like how iPad dethroned Windows PCs for average home user but not Mac because Windows had the monopoly and then an innovation destroyed MS in this space and not a competitor.

I don't think Google dethrones Yahoo and AltaVista scenario will occur again.

Hi,

Interesting idea. Definitely see an overlap with eReader markets and looking at text only contents.

How does it work?

It ignores pages on which it detects frameworks for ui and ads or any javascript code at all?

It's the problem of diminishing returns.

If one is at rock bottom then working hard and being productive can get them to middle class lifestyle. It works. Helped billions of people in the past few decades.

But starting from middle class and working hard won't make riches. Think of it physically. A hardworking person can build a house compared to a drunkard who will be homeless. Yet the same hard working person can't build million houses and get insanely wealthy.

To get truly materially rich (millions+ usd, servants, yachts, etc) one usually needs to be evil and screw over other people. Productivity, in the sense of a machine making houses in the millions, would make the inventor fairly rich. But this is the exception rather than the rule. Most riches are arrived at immorally as parent comment mentions.

> Is nationalisation evil in itself?

Oh very much it is. Come to the lovely Eastern Europe and see for yourself how good national owned companies are. Full of useless bureaucrats put there to ensure voters so the ruling party can continue ruling. Also in huge debts which are paid by more taxes so the working people pay for the lazy.

Unless the human race somehow chains itself to selflessness, nationalization + democracy is a sure way to destroy any organization. Now privately owned is not much better but in theory can be replaced with a competitor. Not so much for a national organization.

Source: Living and suffering daily in Eastern Europe.

That won't stop organizations with lots of money unless the cost for more credits raises exponentially. Everyone will have to stop when millions and billions get into play.

Inconvenience. With the second account one has to re-friend all the users from the first account. Also it would lead to bad social standing as by re-friending it will be obvious they are breaking the 100 posts per person per life rule. This could even lead to automatic bans by studying the connection structure.

It could be the case that they 'hire' someone internally from role A to the advertised role B. But due to some legal requirements they have to run an Ad so they appear open to all candidates.

I had pretty good experience with recruiting agents. Sadly didn't get the job as I flunked the leetcoding part but I got an interview compared to rejection when I applied via forms and CV.

Ford was paying more money to employees who were more conformist to his standard of ideal living.

To the hyper individualistic culture like USA today this can only seem bad but this is Japan with factory towns [1] and generally more collectivist culture.

This also need not be bad. Clean and simple lifestyles could make many more people healthier and happier compared to constant analysis-paralysis state of choice. One could even say that this company is morally better than other companies due to promoting a lifestyle which makes people happier in the long term.

[1] https://www.toyota-global.com/company/history_of_toyota/75ye...

> Finally, is the challenge of having a society built around strong ethic ties. Even if Helsinki were to switch to English, there will be always be outsiders i.e. those who are not ethnic Finns. Again, this is not something that policy can solve and is one of the reasons that silicon valley and North America in general do very well at.

Recently I read a book called Persian Fire by Tom Holland. One detail which struck me as very interesting was how the first democratic ruler of Athens, Cleisthenes, solved the tribalism problem. Here tribalism is the perfect word as the average Athentian citizen of that time associated with one of the 10 tribes based on surnames. His solution was to invent 150 demesnes and have people arbitrarily assigned to them. He also invented new surnames.

Something similar is happening in contemporary Singapore. Every neighbourhood must not have more than X% of any ethnic group.

My point is that policy can solve this. Might make many people unhappy, like Finns who are forbidden to live in some zip codes because there are too many Finns, but policy is definitely a way to approach this problem (if it is a problem, nothing wrong with mono-cultural societies who do not want foreigners).

Can speak and will speak are two different things.

Many people can speak English but would like not to. This is especially the case in groups where foreigners are much less in number than natives. They can all speak English but feel unhappy about having to do that to accommodate the foreigners.

Unless Helsinki citizens reach a state in which English is a native language for them and feel other English speakers as co-natives, this whole declaration will be essentially useless at the level of the people foreigners interact with every day. And these matter most. Shopkeeps, colleagues, etc.

In essence being able to speak English does not make one less tribal. Cultural integration of foreigners is very hard to do without learning and living with the native language. Due to this 'expats' flock to places where they pay least taxes and have largest amount of other foreigners to socialize with, e.g. Amsterdam.

It's not that simple because that way you lose too much information. Actually it is more likely the whole system would fail if weights at any layer are reset.

There is a way to selectively unlearn something via Memory Aware Synapses (MAS): - https://arxiv.org/abs/1711.09601

The idea was developed mostly for transfer learning as in learn new stuff on a new domain but do not forget the old stuff as well. For forgetting it could be trained on some old images + all zeros target mask and the MAS to preserve everything else.

Bingo. All of Apple's share price and consequently of Tim Cook's income here comes from people paying money to Apple and boosting the earnings. Don't like Tim Cook earning a lot of money => Don't buy any Apple products or services. Simple as that.

If I had $750m (or any more sizable wealth really, even 10m would be enough for where I live), would be to improve wealth inequality from the roots via the boring and slow route. Better education for children (equipment, teachers, etc), regular food on the table (big problem for many people), counselings and medicine for abusive parents.

Second best thing IMO would be to start some business which is sustainable and doing some social good. For example making e-readers and cheap e-books for the aforementioned children such that their parents can actually buy them. Side note: e-readers are discounted even today (selling at loss) and most money comes from e-Book sales.

Speaking of Tesla's share price. It has Price/Earnings (PE) ratio of 1095 while a normal ratio is around 5 to 10, perhaps around 20 for tech companies. Apple's PE is 29.

Share prices have definitely lost all sense in March 2020. One should keep that in mind when analysing the current economy.

Why there should be?

We live in a global free trade era. Every entity buys cheapest and sells dearest as they can to the best of the information they have.

Labour is just one more thing available for buying.

Not saying this is the best possible humans can do but in the current world economic system it makes no sense for anyone to dis incentivise employing cheaper foreign labour.

That's the biggest problem with the current monopoly capitalism. Centralization of power into single people making decisions with momentous impacts (be they 140+ IQ or total idiots as some other CEOs are). A good economy is made from small and medium enterprises in which some of them failing do not make much of a difference. That is also the basis of a good investment strategy, diversification.

If the single person in power does good then everything is good (more for the top people but good for everyone nonetheless). But if they do bad then thousands to hundreds of thousands of people suffer. (Reminds you of monarchy?) And the current way of looking at this is just to hope that the person in power will be happy to do good due to more shares. What would be, in my opinion, better would be to split FAANGS and encourage SMEs (small and medium enterprises)

Cries in InteliJ (java) and 1GB+ ram usage.

The Elm Architecture is not that super novel. Actually it came from game development from the Components way of thinking. Take a look at this book. It's a classic for gamedev: - https://gameprogrammingpatterns.com/

The Artificial Neural Network is essentially just a huge set of formulas involving millions and sometimes even billions of values. Debugging in conventional software engineering sense of having a breakpoint and looking at values does not work.

The current state of the art for analysis is ShAP: - https://github.com/slundberg/shap

ShAP is primarily an instance based explainer (one image = one explanation) but if you run it over multiple instances it is possible to gather global model insights on the data. The internals of the model are still quite unexplainable compared to decision trees or anything a human can code (horrible code aside).

There is a group at ETH doing work on adversarial attacks: - https://www.sri.inf.ethz.ch/publications/

While not directly related to explainability their work is on providing bounds on how much can corruption of the input still provide valid output. Very interesting and practically relevant as well.

Finally there is also common sense. If race was a large factor in prediction then the model will implicitly learn to predict races. I am not in medicine and do not know how much it is but if it is then the only way to not learn race prediction is to make race not correlated to the targets.

To do anything software dev you need some basic level of smarts. People who have them can also be trained/educated in any other scientific disciplines and create value there. Talent is a zero sum game. If one works on photo apps then that same person does not work on mining tech.

The main selling point of Uber/Lyft are the network effects (same app working everywhere) and the ability to judge drivers by their rating. Price IMO is not that important because people who are price-conscious drive their own vehicles. Cab companies are too fragmented to get something like such a system running.

This ironically shows how size matters a lot. Be it state investment or huge VC, more cash, more power and more centralization means better service (if the megamoney was spent well).

I agree that Uber's way of underpaying drivers is not scalable (basic mathematics) and price hikes should happen to make the numbers work out. However the service level now people expect is very much innovative.

I wish people who dislike Uber as a service lived in places with bad cabs and bad public transit. Uber is a blessing for the user in such areas. Pricewise it is sketchy and likely unethical but on the service level it is 10x and well worth of existing. I am not saying that Uber as employer is good, but damn the service level blew me off first time I went to another city that had Uber.

The bigger you are the stronger is the pull of the average.

Could you share what python libraries you use in your projects for ML and NLP?

Not a single-person company but I worked on projects and in roles where I was the single person making decisions. Here is my small overview of the ML (machine learning) and DS (data science) world. Hope you find it useful

> What is your tech stack?

The choices are python, scala, julia, matlab and R. Of these python takes at least 90% of the market. I am using python and will do so for the forseable future. Don't like it but it's super practical thing to do. Fingers crossed Julia hits mainstream in the next few years.

Since I am mostly in the computer vision world my stack is torch, torchvision, pytorch-lighting, opencv, scikit-image, scipy and numpy. From time to time i call upon numba when it is impossible to avoid a loop (python loops are sloooooow, numba JIT compiles). Very rarely i coded stuff in C++ and used cython to link it to rest of python. Most of the stuff is already there. For model inference serving i use fast api.

> Why did you choose it?

I chose python because there is no real choice here. It's a monopoly. Winner with most ready to use packages takes all. I am rooting for Julia to eventually win over python as using python as the (big) data language is contradictory given how slow the interpreter is. Also it is not a GPU native language.

Torch was more of a personal choice as I really hated tensorflow 1.X versions and the compute graph. It was hard to debug and follow what is going on. Tensorflow 2.X moved to same mode of computation like torch but it was already too late for me.

> Do you think your choices had any impact on your success?

I think going with tensorflow would have worked as well. Lot more annoying but good base. Using language other than python would have been a bad idea. Practical ML is a lot like JavaScript WebDev. Get a bunch of stuff from 3rd party packages and write some glue code.

Words change their meaning as languages evolve. Nothing unusual. I can share this experience that in a more professional setting hacky code is bad code.

But originally back in the 80s and 90s and even the 2000s a hacker was someone who did something different with their devices just for the sake of it. It was a way of interacting with technology. Finding new uses for technology, finding ways to break it. Pure exploration. From the 2000s with massive expansion of personal computing hacker culture died down. No one wants to be associated with ransomware and silk road. Also cryptocurrencies and all the speculations with them made a huge disservice to the hacker community.

It is not 100% dead tho. There is probably some sort of a community hacklab in the place where you live. Give it a visit (after corona dies down a bit).

They can also be a hazard as they can crash much easier during a storm, earthquake or hurricane. Especially if they are dry and not alive any more.

IMO solving overheated cities is not just more trees but less population density. Less buildings, less concrete, less asphalt. Less generation of heat and pollution in the first place. Less is more essentially.

It's a structural issue caused by the way wealth creation works for majority of people in tech. Job hopping, trendy frameworks in CV, "high-impact" projects done ASAP, etc.

No one wants to do boring, slow pace work with lots of planning, reflection and introspection. And why would they do it? These kind of jobs are usually worst paid. We, the practitioners, have every economic incentive to go the other route.

The problem goes far wider in tech than just ML. And unless the society collectively learns to appreciate patience and long-term thinking, as virtues above all else, it won't go away any time soon. What can be done is to discourage use of ML systems if an explainable deterministic system can be used (even one developed in a rush). For example credit scoring. Rules are good while black box artificial neural network isn't, even if the NN has some % more accuracy. Then if the rules are not good then can be amended and in special cases customer support could also override the rules based on human (hopefully unbiased) judgement.

The problem mentioned in the article of COVID-19 detection based on radiology scans is an example of a system which needs ANNs due to the nature of image processing (very difficult problem for rules AI). While techniques such as ShAP could be helpful a radiologist still needs to check because ANNs learn a lot of useless noise very often and the prediction can be nonsensical. Here it would be best to use PCR tests, serology or any more traditional and "boring" tool as it works. Luckily that is the case and shit CNN models start and end their lives in some useless paper.

The central idea of the black swan is that impossible events, as in 6+ standard deviations from the mean in a Gaussian setting, are far more probable because the underlying distribution was not Gaussian but one similar to a power law distribution.

The reason why these events happen is because humans simplify the world too much. I am an European. I see white swans my entire life. Hence black swans must not exist. Go to Australia. See black swans. Be shocked that statistics are essentially lies.

Our (humanity's) current understanding of gold is that it is scarce, hard to produce and it is valued by other humans. Thus under this model gold going to 100000x or 0.00001x of its current price is next to impossible. If it did we would think it is a black swan event.

But let's say someone out there finally learned the secret of alchemy and can flood the market with lots of gold making not scarce any more. For them the price tanking to 0.000001x is not a black swan but just business because they know the underlying reality. Also conversely if it turns out that only 0.0001% of the current gold supply is real while other is tainted in some way the price of the real could spike.

It is all about information asymmetry (human vs human) and naive simplification (human vs nature).

People usually buy gold to hedge against inflation of currencies. In a black swan case of alchemy this can be really bad but I guess someone at Palantir probably checked that the gold scarcity and demand is still a valid idea. Or not. People assumed the banks are good and we got 2007. Can't trust anyone.