Iv

When mainstream finance magazines tell you what is the latest expensive "must-have" item to a party some still believe to be the ultimate alternative gathering.

Actually Slashdot's articles are more relevant to my interests now!

Because of that I had moved away a few years ago but a friend pointed to me that it is still worthwhile to scan 3 pages of links for the one relevant actual tech bit.

Maybe I should check out slashdot again...

American dystopian movies always give a chance to the hero because a good ending is like an unviolable law of the Hollywood universe.

In reality, automated killing happens without the target receiving previous warning.

Real story: I needed to register an instagram account a few months ago. I tried my nick "Iv", of course it was unavailable. I tried many different alternatives and finally managed to get "Iv but the real Iv" or ivbuttherealiv. Recently I saw my account was banned. I then realized that there was a "butt" in there.

Now I will never have an IG account linked to my FB. Oh well, I can live without that but thank god I don't depend on that platform for business. It made me laugh but there is zero appeal available.

No, it is missing several important information.

<Obi Wan's voice> "OCaml... That's a name I haven't heard in a long time..."

If a profitable company is making a loan it does not need to a subsidiary it controls, that's a fictive loan used to disguise a transfer.

When licensing costs are paid to a shell company with 3 employees in a country that is neither the place where the HQ are or where the company originates from, that's fictive.

Maybe it takes PhD to develop that through the Microsoft development process.

It is probably a reference to the SDI program, also called the Starwars program. Note that the economic impact on USSR is just speculation.

https://en.m.wikipedia.org/wiki/Strategic_Defense_Initiative...

It is time to cut down the tax havens from international financial circuits.

That's like saying murder should be legal as it is impossible to totally prevent it.

Several EU countries (France and UK) have, IIRC, made very clear that fictive debt or fictive licensing fees are fraud an would be judged as such.

You can't just throw more compute at an algorithm and have it spit better results! Well, unless your algorithm uses deep learning that is. And that's why teams with a lot of compute use DL, and why team that use DL use a lot of compute

Fixed title: Sparta was NOT an army of super warrior.

From what I understand, it was a fantasy model heralded by conservative Greeks as an example for all but is as much based in reality than the golden past of the people calling to "make America great again". It is projecting into the past an ideal that never was.

I guess that's part of the Dilbert principle: organizations grow until they outgrow the organizational capabilities of their managers.

"So what are these rituals for?"

"Some are purely ritual, but some are there so that we can guard this dimension from horrors that live within!"

"Uh, I guess we should at least do the latter?"

"Oh? You can tell the difference? That's great!"

-- Girl Genius

GPT-3 stays incredibly in-topic. You can give it logic problems it won't be able to solve, but GPT-3 is capable of things that, as a GOFAI supporter, I never thought brute-force connexionist approach could do. I am now very cautious before saying a task can't be done by DL because it requires "understanding".

These approaches discover concepts and the relationship between them, and use that in their tasks. It is not far-fetched to say that there is some kind of understanding there.

For now we have trained it to generate fake text and basically made a master bullshitter, but I have no doubt that it can easily extract meaning and intent from text.

There are several datasets of programming puzzles, of increasing difficulty.

Basically, you write a test, let the algorithm find the program that passes it. Hopefully, at one point you reach GPT-3 level performances where it is able to imagine programs for tests it never saw.

They brake only when they stop at the station. If you were sleeping next to the tracks but not next to a station, you probably did not hear them much.

Just like the first Go playing programs. Just you wait until it is trained by "self-play" against a compiler.

Huge amounts of data was necessary in the early models and is still when you want to win DL competition.

I fine tuned YOLOv5 with a few dozens hand-labelled images to make an object detector in a semi-controlled environment.

The idea that you need a million images to train a detector or a classifier is now totally wrong. Fine-tuning can be done on a very small dataset.

Well the main sound you hear when a train arrives in a station is the sound of brakes. Its frequency and volume changes as the train slows down. You'll need to analyze the physics of that before extracting doppler shift from it.

Also, depending on the track used, there may be trains passing by without braking, so you will need at least a classifier to sort these two cases.

I'd argue that using ML to build such a classifier is almost always a time saver.

And if you have the ML pipeline there, why not try to train it to recognize the speed while we are at it? It will likely find out about doppler shift but also do things that would take ages to code manually:

- Use volume levels and volume level differences - Use the clicks at rails junctions to evaluate the speed - Recognize the intensity of the braking/engine running - Use cues like rails vibration at certain speed - Adjust for air pressure difference when it hears the rain

All of that for free. Nowadays, going ML first is becoming a pretty good idea actually.

I went into ML when I realized that this piece of advice is now wrong, at least in computer vision.

It was a few years ago. I had to classify pictures of closed and opened hands. I thought surely I don't need ML for simple stuff like that: a hue filter, a blob detector, a perimeter/area ratio should give me a first prototype faster and given the little amount of data I had (about a hundred images of each), not worth the headache. I quickly had a simple detector with 80% success rate.

Then as I was learning a new ML framework, I tried it too, thinking that would surely be overengineering for a poor result. I took the VGG16 cat-or-dog sample, replaced the training set with my poorly scaled, non-normalized one, ran training for a few hours and, yes, outperformed the simple detector that took me much longer to write.

Now in computer vision, I think it makes sense to try ML first, and if you are doing common tasks like classification or localization of objects, setting up a prototype with pre-trained models has become ridiculously easy. Try that first, and then try to outperform that simple baseline. In most case, it will be hard and instead worth improving the ML way.

Yes but at that point you may simply use classic white paint I guess. Also one thing to consider is that going front 90% to 80% may not look like a big difference but that's still doubling the heat that comes in

I see CG, I don't see an actual product. Does it exist yet?

In France taxes come pre-filled. Cheating is equally simple: you just "forget" to fill additional fields.

No, TurboTax seems like a clear case of lobbying where millions are inconvenience just so that a handful of executive keep an underserved revenue.

I'd love that a visit to a studio to see how these images are made be part of the mandatory school curriculum.

At this point I am skeptical of either prediction, just like remote work, homeschooling could gain many adepts.

Then it is called journalism. It already exists and it is usually considered unethical to fund it when you are one of the object of their scrutiny.