> Writing reviews isn’t, like, a test or anything. You don’t get graded on it. So I think it is wrong to think of this tool as cheating.

Except that since last year, it kind of is. It is now mandatory for some large conferences (such as CVPR) for authors to do reviews if they submit a paper. Failure to review, or reviews that are neglectful, can lead to a desk reject of their submission.

Seems like… a pretty bad policy, right? I wouldn’t mind doing a mandatory review, but I certainly wouldn’t want to be on the receiving end of some grumpy person who didn’t want to do one.

Even the US is a good counterexample. It is a widely fragmented country, yet that worked for decades without degrading into a totalitarian state.

It has been back for a while in the form of quantum fields.

It's more consistent.

As a rule of thumb, Bloom filters require 10 bits per element for a 1% false positive rate. Say 100 pixels changed between the frames, that 1000 bits or ~125 bytes to store the which-pixel-changed-map.

Runlength encoding of the (in)active bits can use anything from ~10-200 bytes for 100 pixels (Say 1 byte per run, 200 runs of active vs. inactive in the worst case).

> What contributed more towards success in my opinion are "shortcut connections" through layers which enable more influence on early layers during learning.

For those who don't know, that is the idea behind ResNet (He et al., Deep Residual Learning for Image Recognition, https://arxiv.org/abs/1512.03385), one of the most influential papers in deep learning of all time.

Residual connections make it possible to train networks that are arbitrarily deep. Before ResNet, networks that were too deep were essentially not trainable due to vanishing or exploding gradients.

It's really nice to have your personal intuitions in a field you barely know confirmed by research.

Do you have an example? That just sounds like a hash that is a few bits longer.

Mostly use of GCM (Galois/Counter Mode). Usually you tag the key, but you can also tag the value to check verification of collisions instead.

But as I said, slow.

There is of course a slight chicken-egg-thing here: If there was no (dynamic) branch prediction, we (as in compilers) would emit different code that is faster for non-predicting CPUs (and presumably slower for predicting CPUs). That would mitigate a bit of that 10x.

A bit. I think we've shown time and time again that letting the compiler do what the CPU is doing doesn't work out, most recently with Itanium.

In theory, sure. In practice, complex technological and political issues remain - apparent by the fact that no country has solved the issue yet.

You apparently stable salt mines start leaking. Locals don't like having toxic stuff buried below them. Other countries dislike that you dump nuclear waste in the middle of the Atlantic. Digging deep becomes too expensive.

While you are fully right, that is very US-centric. Tuition is practically free in large parts of Europe, for example.

"Missing" information is a huge red flag as well, though. If an Arbeitszeugnis only lists the dates and what the person worked on, and nothing about results or behavior, we would never invite them for an interview.

It could be that the employee left due to being on bad terms with the employer, for something at fault with the employer. I personally wouldnt take a basic one as that hard a strike.