vhold

I really want to understand how Amazon doesn’t understand this. What is it that has them so totally blind to what everybody else can so plainly see?

I wonder how much it would have cost to simply recreate Google Earth VR, which is still one of the most compelling VR applications, period. It was released 6 years ago.

The most likely reason would be to neutralize us before we can become a threat far far in the future. Subjugation could come in subtler forms, like in the book The Three Body problem where the aliens halt human scientific progress by stealthily interfering with science.

One camera produces millions of bytes of data every single frame, an ultrasonic sensor is useful producing just 1 or 2 bytes of data in the same time span. (distance to something within the sensor's cone).

So it seems like a totally ridiculous argument that ultrasonic sensors create some kind of data processing overload.

An ultrasonic sensor makes it possible to implement incredibly simple and reliable safety features with well known performance characteristics. Processing an image with ML to produce the same effect has tons of edge cases where it might not work, and nobody knows when it won't work, and every update to the system could introduce regressions.

It's why they had to disable certain features when they got rid of the ultrasonic sensors. Those features may come back some day, but I bet they'll never be as reliable, and certainly won't be as predictable.

https://www.pcmag.com/news/tesla-removes-ultrasonic-sensors-...

Comes from things like: https://www.lexaloffle.com/pico-8.php

More here: https://github.com/paladin-t/fantasy

Some very famous paintings it can almost reproduce, like the Mona Lisa and The Last Supper. It can’t get it quite right but I think it’s close enough to be considered copying. So there might be a copyrighted instance of that, but I haven’t seen one yet.

All of this with 207 kB over the network. The two slowest requests were to google analytics.

If true, it reminds me of the reputation that Jack Tramiel, founder of Commodore computers, had in the 80s of screwing over his suppliers to the point they would no longer do business with him.

I wouldn't downvote you, but I think you should be aware that meta-analysis is making the case that homeopathy is better than placebo and has an unknown mechanism of working, which is a much bolder claim than you originally made.

I disagree with the results of the risk of bias assessments of the studies involved in that meta-analysis: https://www.nature.com/articles/s41390-022-02127-3/tables/1

Out of the 6 studies used, only 2 are from general research institutions. The other 4 are all specialists in homeopathy or private practices using homeopathy. Seems like the risk of bias there would be extremely high.

For those who have not seen the Walmart VR demo, it's really something else. Something is going very very wrong at Meta if they thought this looked enticing:

https://www.youtube.com/watch?v=uLcaDStxljw

To me the difference is this: https://news.ycombinator.com/item?id=27710287

It's possible for generation models to perfectly memorize and reproduce training data, at which point I view it as a sort of indexed slightly-lossy compression, but it's almost never happening with image generation because the models are too small to memorize billions of pictures, it can't produce copies.

Stable diffusion 1.4 has been shrunk to around 4.3GB, and has around 900 million parameters.

I don't know how big Copilot is, but a relatively recently released 20 billion parameter language model is over 40GB. ( https://huggingface.co/EleutherAI/gpt-neox-20b/tree/main ) GPT-3, according to OpenAI, is 175 billion parameters.

It's possible there are some images in there you can pull out exactly as is from the training data, if they were to appear enough times, like I suspect the Mona Lisa could be almost identically reconstructed, but it would take a lot of random generation. I'm trying it now and most of the images are cropped, colors blown out, wrong number of hands or fingers, eyes are wrong, etc.

The lawsuit also says that Teledyne knew that the DRS had a much higher failure rate than advertised and didn't tell the Air Force.

I think that's exactly how it would have gone done except for the fact Teledyne is accused of replacing the counterfeit parts when the board was sent to them for inspection, which would almost certainly prove they knew they were fake.

I get the impression that John Carmack has achieved this, and I bet it required him to forcefully take control of his schedule.

Which reminds me of this game: https://www.saynomo.re/

I think VR needs pretty badly to pivot in general, it's an amazing technology that tries to replace phones/computer interfaces where it should be working with them.

I don't want to strap a thing to my face and "be in VR" for a long period of time, navigating menus by pressing giant virtual buttons or pointing lasers at them. I want to, while using google maps on a phone or computer, be able to hold the headset up to my face and look around briefly and then take it back off. Same with data visualizations, 3D models, short-form entertainment, etc.

Standing in VR looking at loading screens is an awful experience. It should already be loaded to what I want to do.

It's hard to imagine Meta making that pivot since they seem to want to have a walled garden like Apple's. But they could at the very least make navigating/loading by a phone app an option. Even track the phone in VR, have a mirrored twin of it (with a kind of blown-up holographic version of it you can more easily read), that you can use in and outside of VR, and instead of their controllers.

The ruling mentions "There is little caselaw interpreting this statutory definition." in regard to that question. It mentions one case where building 8 residential units was a "housing development project"

It then goes into the fact that the law in question refers to a "housing development project" in one of 3 ways, 2 of which definitely don't apply, leaving behind the phrase "Residential units only", which they then point "units" is plural, not singular.

So perhaps.. 2 residential units would have been a housing development project.

Here is the resolution they adopted: https://sfgov.legistar.com/LegislationDetail.aspx?ID=5742708...

https://sfgov.legistar.com/View.ashx?M=F&ID=11095427&GUID=46...

I think it is useful because it provides a fairly compelling and surprisingly specific list of justifications.

It's really nuts, computer programming has this gigantic language translation component to it, so asking somebody to translate from English, to another human language, then to a computer language, is fraught full of peril and basically fails most of the time.

The only teams that can pull it off have people who are really talented in the company's native language, and computer programming, and communicating with non-native language people. Those people are super rare and are not paid nearly enough to encourage more people to seek out that niche.

Reminds me of how my Windows Mobile phone over 20 years ago would sometimes bring up the dialpad without the "9" available, which was the last straw for me and got me to switch.

I think it's a genetic trait diversity bottleneck? Mammals almost all went in the direction of nocturnal adaption and a lot of adaptions useful during the day were lost over time.

My hunch is that is the result of this: https://github.com/CompVis/stable-diffusion#weights

> 515k steps at resolution 512x512 on "laion-improved-aesthetics" (a subset of laion2B-en, filtered to images with an original size >= 512x512, estimated aesthetics score > 5.0

https://github.com/LAION-AI/laion-datasets/blob/main/laion-a... for more details.

What's remarkable is this: https://github.com/LAION-AI/laion-datasets/blob/main/laion-a...

That aesthetic predictor was apparently trained on only 4000 images. If my thinking is correct, imagine the impact those 4000 ratings have had on all of the output of this model.

You can see samples (some NSFW) of different images from the original training set in different rating buckets here, to get an idea of what was included or not in those training steps. http://3080.rom1504.fr/aesthetic/aesthetic_viz.html

More arguments to always use HTTPS:

https://https.cio.gov/everything/

https://developers.google.com/search/blog/2014/08/https-as-r...

The practical answer is that you go and look at a project's release history at its primary repo. I don't know of a good way to browse all apache projects by livelihood though. The closest thing I know of to keep track of Apache projects overall is this:

https://lists.apache.org/list.html?announce@apache.org

But just eyeballing that really quick I can see it's not authoritative, because a new version of Apache Spark just came out and isn't announced there.

https://spark.apache.org/news/index.html

There is also this, but it's very cluttered with submodules, but it does appear more authoritative.

https://projects.apache.org/releases.html

AlertManager is one component of a more complicated infrastructure.

https://prometheus.io/docs/introduction/overview/#architectu...

https://kubernetes.io/docs/concepts/overview/components/

The image of the email he did post says he received previous emails with the reason.

See also: https://github.com/nmap/nmap/issues/2237

Another possibility is that people were using it for money laundering.

One method is to create lower-than-shovelware games, buy them with shadily acquired cryptocurrency, stolen credit cards or scammed gift cards and take real money payment from Valve. Egregious example: https://mmofallout.com/valve-bans-yet-another-laundering-gam...

With synthetic data you can also generate the thing you are trying to solve with the AI in the first place, like generating a depth map, and an object classification map to go with a simulated image.

With real world data a human will have to label it.

I don't think they even realize that is what happened. You can see the flawed thinking throughout the entire description of the experiment. They anthropomorphized websites, imagining them to have the abilities that the humans behind them have.

That's probably the big thing that people should learn from this, I think this is a pretty common misconception.

One of the most interesting technical details is buried in a giant quote.

> All it took was a little bit of collaboration and communication, and all of a sudden, grip and electric are moving generators a hundred yards away instead of having them right around the corner from the set.