jrd79

I believe resistive losses are the primary limiting factor, not insulation.

Is it possible to collect actual observations to confirm or deny the results of this modeling? A key part of any well formed scientific hypothesis is that it can be validated or disproved by real world experimentation or observation (falsifiable). Where will that data come from in this case? Computer models are an important part of science, but if they can't be validated against data, what's the point?

Another competitor that has AFAIK the best unit pricing is: https://pixian.ai

(I'm a dev on the project.) We have another background-removal service, ClippingMagic.com, that is built around an editor to let you fix the errors in the automatic result. You may want to give it a try for your mineral photos.

(I'm a dev.) Not everyone has macOS, and running a DL model in your browser is not exactly mainstream. Which is why segment leader remove.bg gets an estimated 35M in monthly organic traffic. Also, we believe our results are significantly better than those offered by the open-source models we've seen.

(I'm a dev on the project.) It is free while in beta. We've not implemented pricing yet, as we are a small team and just shipped the latest version of the model.

(I'm a dev on the project.) We have no plans to jack up prices. We see the whole market moving to low-margin cost+ pricing and we want to lead rather than follow. If we raised prices later, we'd only expose ourselves to being disrupted in the way that we are hoping to disrupt right now. Low margin plays are all about operational efficiencies, so that we can turn a profit at price points that other providers cannot. That is our laser focus, which is why our processing time is so quick.

Clipping Magic is a totally different product. It is editor-based with a bunch of post-clip effects and features. The editor allows you to fix errors in a way that a single-shot DL-based solution simply does not. We don't actually see the services as competing with each other, since DL-based solutions have taken over the portion of the market where 80-95% success rate and some errors are ok, so long as it is fast and cheap.

(I'm a dev on the project.) The privacy policy is an old and generic one that we use across a bunch of sites. It should be updated. Our retention policy on this site is as stated on the front page FAQ. After five days, the records are deleted.

(I'm a dev on the project.) We've not decided on the exact term of the credits, but they will be long-lasting, so you can pay $5 for 250 images and use that over the course of a few years. We'd make them non-expiring, but that creates an unbounded liability.

In the third paragraph, the author writes that COVID was probably "cooked up in the kitchen of nature", and links to an article to support that assertion. But the linked-to article does not make that point. In the comments section of the article, a commenter pointed this out, and the author replied to that specific comment with an answer that did not address the issue but expressed the author's personal opinion on the matter, and has not updated the article to be more clear.

I'm the other main developer on this project. We see this as part of a broader trend where the margins are going to shrink dramatically for those selling AI tools online. Rather than fight that and come in with prices only a bit below our main competitors, we have decided to just jump straight to the eventual end state and price our services as cost-plus.

This means we are dramatically cheaper than our competitors. It is not even close. But we have solid quality and great dev ops.

The broad development is excellent news for consumers of this technology, and is the inevitable future anyway.

I deleted all social media over the span from 2016 to 2019, and after an short period of actively missing it, I don't even think about it anymore, unless the topic comes up as one of societal concern, or in connection with my kids. Phones are still distracting without social media, but materially less so. Being off social media has not just been a net positive, it is almost an unmitigated positive. The social connection it promises to provide is illusory, and it gets in the way of actual social interactions. It is sad to see people who obviously know each other sitting next to each other while staring at their phones, many of them presumable looking at social media. I see it among young people at my kids' events, among my social set, and especially among other families out to dinner.

The harm it appears to be causing to kids is my main concern, given social pressure to do what other kids are doing. Banning communication platforms for kids would be hard to implement in practice. A good starting point would be to ban phones from K-12 schools. They are raising a generation of kids with really bad habits and poorly developed social skills, attention spans, and creativity.

I'm the other main dev (and co-founder) on this project. Some interesting things to note:

* This is all our data from our other site, so it is very real-world and diverse. We don't have a short list of supported categories like some of our competitors seem to use, but a single monolithic network that handles all inputs.

* We are really focused on speed and efficiency, so this is using a fairly small network compared to the huge networks that have been coming out in recent years. This keeps (our) costs and runtimes down, so the site is fast and we can offer more for free than our competitors without it hurting too much.

* This market (for background removal) has stagnated a bit, with the top couple players seeming a bit entrenched. This means there has been a lot less price competition than I would have expected, given how many upstarts there are. Unclear why that is, but we're hoping to shake things up a little.

Regarding all the anecdotes, I would have said the same thing 8 years ago when I moved to my current house in suburban Minneapolis. Then we had some landscaping done and added a bunch of planting beds to the otherwise boring grass-only yard. And my kids threw some milkweed seeds on a slope that is too steep to mow. Now we've got tons of bumblebees, honey bees, monarch butterflies, and other insects. I think the range of most insects is quite small, so maybe people who miss the bugs just need to plant a bit more in their yards :).

Second point: the windshield splats proxy is too confounded to be useful. Use actual measurements or reasonable people will discount the claims (as I have done).

Do you have an example of this being done on a large scale? Sincere question. I've just never heard of it being done.

Modern technology makes it possible to build cities in arid environments, but it doesn't make it wise.

Carbon dioxide was not considered a pollutant when the law was passed, so congress should specifically authorize such a significant expansion of the EPA authority. Rule by executive fiat is never wise, and if the case were clear cut on this matter, it would not be so hard to get congress to act. There isn't and never was democratic consensus on this topic, and people who support it need to convince people who do not before it can become public policy.

Shameless plug: weatherspark.com has a great compare feature that is extremely useful when evaluating the relative merits of a given location's climate. See the following comparison page between Algiers and LA:

https://weatherspark.com/compare/y/48929~1705/Comparison-of-...

That is an amazing bug.

The (s)RGB color model is not some legacy choice. It is closely tied to how colors are stored, composited, and communicated to physical displays. And this problem seems a bit artificial - how often do you see a linear gradient between colors like the ones shown in the examples? Finally, it is also trivially easy to get a gradient that looks like the "correct" ones presented in this article, by just adding a few additional color stops along the way that are closer to where you want the intermediate values to be.

Two extra letters. Numbers matter :).

That sounds like it applies to websites of businesses with physical locations, no?

From what I can tell, Title III (public accommodations) only applies to businesses with physical locations that serve the public. So are purely digital Saas businesses not subject to this lawsuit risk? If that is not the case, can someone more knowledgeable on the subject post a link to the relevant section of law or official regulation?

The Wikipedia page states: "The Académie is France's official authority on the usages, vocabulary, and grammar of the French language". And "The body has the duty of acting as an official authority on the language."

So while they may lack the power to enforce their official positions, the mere existence of an official state-sanctioned authority for a language is a notable difference from English.

You are avoiding the question of whether it is appropriate to present the results of a linear regression on data that is so poorly explained by a linear relationship.

Random looking balls of data points don't have slopes. It is invalid to perform a linear fit on data that does not derive in large part from a linear generative process. And presenting a fit from a model that is facially absurd to apply is bad data science. Whether or not an informed reader would discount the absurd model fit is not material to whether it is appropriate to present such a fit.

They could have binned the data and plotted percentile bands. They could have used a non-parametric density estimator. There are lots of things they could have done to summarize the data and make some sense of the ball of points. But linear regression with slope error bars is not an appropriate choice. That it is easy to compute linear fits, and that it helped them make their point is not justification.

No, I'm complaining that they plotted slope-only 95th percentile error bounds, which is visually deceptive. If they had plotted the vertical spread as well, it would have been clear to the reader that the model they chose to explain the data (a linear regression model), does not fit the data at all. That is also clear from the R^2 values, but that is hidden in a difficult-to-interpret numerical value. So the model is not well suited to modelling the data and all conclusions drawn from that model are unfounded. The slope value and its confidence interval are essentially meaningless because the data is not actually modelled by an affine model, so it is nonsensical to talk about the slope estimate and its uncertainty, as the data is not describable by a slope. Models must fit the data well enough to be plausible in order to be useful aids to understanding the data. These models don't come even close to that standard and should not have been used. Any data scientist worth their salt knows this. The authors either know this and went ahead with it anyway, in which case they are dishonest. Or they don't know this, in which case they should not be using such methods, as their incompetence is made plain for the world to see.

A one-dimensional affine fit (usually called a linear fit) contains two parameters: a slope and an offset. Both have error bounds, and the offset error bounds on this data would be huge. Data presentation that is not intended to deceive would have shown the vertical spread of the estimate too. But that spread would have been so wide that it would reveal that the fit is terrible and that reasonable conclusions cannot be drawn from these model fits. This is not scientific work. It is ideological policy advocacy dressed up as data science.

Those trend line fits are hilarious. Unmodeled confounding factors, anyone? I especially like where they choose if something should be a log-linear fit vs. a linear fit. And the 95% confidence interval only containing less than 10% of the samples. These guys are amazing. And Greg Colburn is a professor! Good job, guys. Go find other work.

Interesting article, but it should be noted that there is no "official name in English" in any meaningful sense of the word. Yes, you can cite what a particular government entity uses, but unlike French, English is not an officially regulated language, and as such, we are all free to call things and places what we please.

Every language has names for foreign places that are not the same as the name that is in local use in that place. It is Munich and Rome, not München and Roma. That is because place names, like all words, are chosen organically by speakers of the language for convenience, ease, and clarity. To an English speaker, "München" or "Kyiv" are hard to pronounce in their native forms, and so we have alternatives. No disrespect is intended, and none should be taken. Foreigners do the same thing for American place names, as well they should.

There is also value in continuity. Changing place names in a language due to external pressure of non-resident non-speakers of the language, places an undue burden on the speakers, for little or no benefit to the speakers of the language. The language is for its speakers, not for non-speakers.

I grew up with "Kiev" and "the Ukraine" (as well as Peking, Bombay, Canton, and many others). It is relatively harmless to change these, which is why most people just go with the flow. But it is not a moral issue, and people who wish to stick with older forms should not be ostracized. No one can stop a scold from being a scold, but most people don't like scolds, even if they sometime acquiesce to their demands out of politeness or a sense of conflict-avoidance.

Where's the live online demo where I can upload something and watch it work? Without that I have no confidence that the system actually works.

[EDIT] Sorry, missed that!