Dzugaru

> you chose to potentially cause more damage

Seriously? I surrounded myself with a metric ton of rigid metal and bazillion safety systems (and pay ridiculous sums for that, mind you) to protect myself and my family in a deadly environment. Cyclist chose to go there practically naked.

And before you say it's not a "deadly environment", it's "deadly people" - that's mostly false. Any human has limits to reaction time, can't see everything and so on, fixing that substantially is really hard if not impossible (self-driving cars, maybe).

Yes, as a driver I basically chose to sign a blood contract to kill people with low probability cause I'm not a perfect machine. But did I have much choice? Driving my mom to a hospital on a bike is impossible. Are you sure you didn't sign other such contracts in you life for yourself and your family? Are you sure your quality of life you chose does not condemn other people to poverty and death?

What would that achieve? If you're late to the blockchain (or really any other distribution channel, like https://truetimestamp.org/ or something like this) your timestamp is greater than original and you own nothing. I don't really know how selling 2nd place works with NFTs but I assume you only can get it from 1st place owner (he has to sign it with his private key), so the scheme is solid. You cannot change a pixel and claim something.

The only problem I see is how the 1st buyer knows the seller is the artist not some random dude who changed 1 pixel (or even didnt). You actually cannot reliably sell something that has been published before cause everyone and their mom can claim ownership before 1st token.

Well, physical art reproductions are high quality already (I have some Van Gogh canvas prints I bought for like 50$ a piece on my walls). So you can't really say traditional art images are scarce too.

The thing is whoever is owns the original in some sense, even if you eventually cannot discern it with a microscope.

I still don't get why this is a scam, there is a SHA-256 hash of the whole image in that token, complete with timestamp and a buyer signature - the buyer can certainly prove he was first (or Nth) to buy this exact thing anytime he wants (if he doesn't lose his private key ofc). Who cares if some URLs included do rot?

It's almost the same idea [0] Galileo used 400 years ago to prove he was first to see Saturn rings by hashing the image description and distributing the hash.

[0] https://en.wikipedia.org/wiki/Rings_of_Saturn

You bootstrap the pos chain onto the pow one, so you have your trivially checkable first part. After that it.s not exactly trivial to fake the whole chain without holding validators private keys (remember, they had stake in pow chain).

Im kinda at a loss about what other vectors of attack possible, hard to find any digestible info, and pos protocol designs are really complex.

It's just natural, that people are salty about proof-of-work that leads to computation devices scarcity. No one asked for this and it's really scary - society can economically decide, that distributed ledger protection is more important, than gaming and science. Just like we economically can decide that global warming is not important.

Good read. Is there a similar take on how proof-of-stake can work (for example current sketch of Ethereum Casper protocol)?

As a programmer I usually try to implement some of the algorithms in language of my choice and throw them at some toy problems (it's good there is no shortage of interesting toy problems in this field). It was really fun with Artificial Intelligence: Modern Approach book (ex: implementing a SAT solver from scratch and solving Minesweeper by pure logic is very satisfying).

Many competitive shooters use ballistics calculations, Battlefield, Apex Legends to name a few.

A great question! I’m still wondering why we all mostly use separate tools for tickets and version control (and knowledge base, and collaboration, and management/hiring, the list goes on), given we already know a lot about software development.

In my experience in computer vision research it doesn’t matter what you use, yes, immediate mode is slightly more convenient, but research time is influenced much more by your computing power, dataset acqusition/labeling/relabeling power and, last but not the least, by code quality and easy and efficient collaboration - thats why you need tools like DVC/Argoproj. We did get amazing results using Caffe v1 back in the day.

Don't know about other countries, but in Russia we say sorry all the time flashing hazard lights briefly. Pretty much every driver on the road knows that.

> Developers can’t easily go back to where they were right

> before an interruption. They need to get into the mindset

> for development and then slowly trace back to where they

> left off. And every fellow developer knows that.

I used to think like that too, but after 10 years programming in many environments, now I'm inclined to disagree. If I cannot refocus immediately after interruption - I know I'm doing something wrong. Maybe the thing is too complex and must be cut down to smaller pieces, maybe I didn't keep enough notes or thinking through - anyhow nowadays I cannot work without a clear understanding of what simple piece of what complex piece I'm doing and where I've written down the other pieces I should do next.

I was sceptical at first but after watching MaNa story video [0] - not anymore. Starcraft2 is a very complex game, but I’ve been able to crush best blizzard AI after a week playing - yes its that weak. Seeing AlphaStar strategy, from simple correct building placement to micro decision making is so much more inspiring. And it was achieved by deep reinforcement learning - never seen a more unstable and unpredictable algorithms in my machine learning career - even in fully observable environments. What DeepMind did - achieving THAT level of play, even if AlphaStar can see the whole map and micro 3 groups of units at once - is incredible.

[0] https://www.youtube.com/watch?v=zgIFoepzhIo

This is why I never ever use size_t over int. I’m a newbie C++ programmer tho, wrote less than 10kloc of C++ code, but my both feet are bleeding cause of exactly this.

My team managed to do pretty cool realtime deep neural net computer vision demos on Jetson TX2, but this beast is an order of magnitude faster and energy efficient, which seems almost unbelievable! We're definitely getting a bunch of those and sticking them to our robot.

I’ve always wanted to find a blog post somewhere about SC2-style swarm pathfinding. Dijkstra part is simple and there are many resources about it, whats hard is avoidance - units shouldn’t walk through each other. Achieving that smooth and correct wave-like motion of zerglings isn’t easy - I’ve tried but the results were ugly.

D or Rust?

Please, don’t ever print screen any code you want to share with someone, it’s really bad for your karma.

Many years of "human optimization" on the same old datasets must lead to severe overfitting, so these results should be taken with a grain of salt. Related research:

https://arxiv.org/abs/1806.00451

Rocket league was meh for me, but then I tried Oculus’s Echo VR zero-gravity soccer - and it’s unbelievably good. Having real physical competitive sport in your room anytime, instantly - it’s like a dream came true.

Not quite watchable on streams and videos currently because first person view on 2d screen is erratic and horribly flat - but I hope games like this is the future.

The argument is still false. You can very well get an advantage from vast amounts of data in similar domains. And more importantly you can have ML insights not possible without it. What if ImageNet was not open to the public? Would we get an AlexNet breakthrough?

Never thought I would see such xenophobic comments on HN. This is really sad.

> The body of evidence to date suggests that you're much more likely to be robbed of your cryptocurrency through fraud or incompetence

Because the technology (right now? maybe we'll see cryptocurrencies built in hardware/os eventually) is for competent geeks, not that greedy people that store their bitcoins on exchange wallets (or use any central server). One has to ask oneself if the blockchain is more reliable for him, just as he does with other things like choosing doctors etc.

> A lot of the blockchain enthusiasts are anarchists or libertarians. They assume society is already failing or on the verge of collapse.

You don't need to assume that to see obvious benefits of a blockchain. As long as there're stupid or evil people in society central authority systems like banks can fail, rob you or be hacked. Blockchain on the other hand is more reliable.

My eyes are bleeding already

I'm so happy not all games are like this, I abhor hidden or unfair mechanics. Even if you sometimes get one-shotted "out of the blue" in, say, Path of Exile, you can be sure game didn't "help" you, there is a crystal clear explanation of every mechanic in Wiki and your 90 lvl hardcore char IS your and your skill achievement.

1) You can't easily test it - bugs can lead to semi-working thing and you may need to run thousands of experiments to average

2) Many hyperparameters - it may be critical to get them right

3) No gradual implementation - it doesn't work until you get all the (often gimmicky) parts right. Take A3C for example - its paper version is parallel, may be hard to implement and debug in your language (and hardware) of choice.

4) Task and reward function choice may be hard - take a look on my question https://stackoverflow.com/questions/44781401/is-openai-gym-c...

Generally, RL algos performance is underwhelming in my experience, without heavy tuning (and so-called "domain expertise" in reward function), but I'm not an expert and OpenAI guys show that you can make the working thing like Dota2 bot, so they give me hope.

What these guys doing is amazing. RL algos are so hard to get right and evaluate - they can be very unstable and depend on subtle details. I've tried it myself (DQN and my version of actor critic) on simple tasks and still don't know if I had errors. Its like you're are chasing a moving target with a neural net - second-order instability :) With this code to pick apart and compare against (especially A2C) it'll be so much easier to hack RL.

> Even though I’m one of the beneficiary of this AI craze, I can’t help but thinking this will burst.

I don't think it will. Level off - maybe.

I've started my work in Computer Vision with classical algorithms (SIFT features, geometry, correlation filters and things alike people were researching for decades). These really worked like garbage, it was a nightmare.

Then we jumped on DL bandwagon - and CV just clicked for me. Now I see it working, not perfectly, not at human level yet, but it works, it's better than everything else and it certainly brings value - not just in CV! Maybe there will be some expectations delayed or even ruined (AGI, fully self-driving cars, dunno), but the tech isn't going anywhere.

At it requires at least some experience and a specific mindset, slightly unusual for a generic programmer. So I don't see a problem with experts, courses, degrees and the like.