Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Egan_BW]

the classic strategy of selling at a loss, but making up in volume?

[methylatedspirit]

Sounds about right. That or it's being combined with an exaggerated version of the Xiaomi strategy as well; plaster ads all over the phone's interface to make up for losses in ad revenue.

[Reelya]

Hey, some good news for a change. Plex is launching their own movie streaming service, with a lot of content they've secured region-free rights to, and it's ad supported so you don't need a subscription. You need the Plex software however.

https://techcrunch.com/2019/12/04/plex-launches-a-free-ad-supported-streaming-service-in-over-200-countries/

[feelotraveller]

Good news? 

As far as I am concerned it is just more of the never-ending commercial stupidity (stupidity of commerce/stupidity of commericals - "it's the stupid economy"  ).  Much better things to do with my life than spend a significant part of it reading advertisements... or supporting content producers of the 'stupid' kind.

But each to their own...

[Reelya]

it's good news because it's another place you can legally get content without a subscription for a whole monthly package. Worst news ever would be them launching yet another subscription service. Enough rival free services might put some of those subscription services out of business sooner. The alternative is a world entirely consisting of walled gardens where you need to subscribe to all of them to get the few bits of content you want from each of them. Ad-based sites have no walls.

As for "stupid" how do you expect a service to afford to keep operating without income? I'd prefer ads on just the content I choose to watch vs having to pay a monthly fee for content I don't even want. Ads are also better feedback, since they only make money off things people actually want to watch, rather than stealing some of your subscription money to produce content you may object to.

[Iduno]

https://www.bbc.com/news/world-us-canada-50667553

Is Elon Musk ever not terrible?

[McTraveller]

Holy cow Windows 10 has terrible multi-monitor scaling.  I tried to connect my laptop to my TV, and Windows scaling is just... nonsensical.

Moving a window from one screen to the other scales the text but the not the window dimensions to match - why doesn't it keep the relative aspect ratio the same?

[dragdeler]

-

[Telgin]

I've honestly never seen that.  DPI differences are what I'd suspect, but I wouldn't expect that to cause text to change independently of other GUI elements.  Do you have text scaling set to something other than 100%?

https://www.bbc.com/news/world-us-canada-50667553

Is Elon Musk ever not terrible?

Wow, this is just now going to court?

[Reelya]

https://www.youtube.com/watch?v=gQddtTdmG_8

Thought I'd share this. It's about making language vectors with neural networks, and how a very simple model leads to some surprisingly cool features.

You start with a vector say 10,000 units long. Each input represents a unique word. You feed it through an NN with a row with 300 units, and the output is also 10,000 units long, and represents the probability of any other word being near the input word. You then train it long enough and on enough language data that it accurately models the language (in a statistical sense).

The real trick here is that your 10000 words are being funneled through a layer only 300 units wide. So those 300 values (a 300 unit long vector) must be compressing down all the information. You can then extract the 300 hidden layer values for any specific word, and do "concept calculus" with them.

And remember, all this is from unsupervised learning from raw language data:

For example, if you take the vector for "king" then subtract the vector for "man" and add the vector for "woman" then you get the vector for "queen". If you take the vector for "London" subtract "England" and add "Japan" you get "Tokyo". take "bark" subtract "dog" and add "cat", you get "miaowing". Also from the data set, take "shirt" subtract "man" and add "woman" and you get "blouse". So the hidden layer cells are clearly encoding a whole bunch of meaningful relationships from the real world in them, at least as far as our encoding of the world in language goes.

[Naturegirl1999]

I am not a coder, but a word with multiple meanings like row with the following meanings for nouns and verbs
Verb in present tense has ing at the end of it. EG. learn(ing)
The past tense has Ed at the end. EG. learn(ed)
Future tense has will before the word. EG. (Will) learn.
You will learn new things. You learned some things before. You are learning now.
Row¹=(noun)
1a: a number of people or things in a more or less straight line.

1b: a line of seats in a theater.

1c: a street with a continuous line of houses along one or both of its sides, especially when specifying houses of a particular type or function.

1d: a horizontal line of entries in a table.

1e: a complete line of stitches in knitting or crochet.

1f: a period of rowing.

Row²
2a: propel (a boat) with oars

2b: travel by propelling a boat with oars

2c: convey (a passenger) in a boat by propelling it with oars

2d: engage in the sport of rowing, especially competitively.

So each of the numbers and letters correspond to meaning. Example:

A row^1a of chairs are on a boat. The driver of the boat is row^2aing across the river.

[Naturegirl1999]

What would a network that did something like my previous post be called? Could it even work?

[Reelya]

Indeed, word vectors are rather fun and useful tools. I say this having worked with them myself. However, there are a few notable drawbacks:

• There is considerable conflation of multiple meanings of a word within a single vector. All of their usages are bottled up into one vector, which can cause problems when trying to use these vectors together to extract the meaning of a sentence. Furthermore, there's no means to suss out the specific meanings of a given word- you have that single vector with them all bottled up in there, and no particularly useful way of extracting the individual meanings that compose it.

A good point and I'm going to focus in on this one rather than write a long post discussing all the details. (mind you I haven't read all the intervening posts here so sorry if I've repeated anything).

One thing that comes to mind to solve or at least heavily mitigate this particular drawback is to do something similar, but you'd need humans to come along and tag every individual word in a training set according to it's actual semantic meaning. There could be multiple meanings per word, but synonyms could actually be condensed down so the total number of words being converted to wouldn't necessarily be a lot larger. The drawback here of course is that you'd need to have enough properly labeled training data. But after that, you could get the algorithm to read raw text, and it could do an semantics-inferring first pass before putting it through the same unsupervised vectorizing algorithm and get much better results.

Another possibility however that wouldn't need manually labeling of all words is to divide nearby-words into sets. i.e. you can design an algorithm to work out if certain sets of words appear near dog, but not near each other. This would strongly imply that the two uses of "dog" were different semantically. So you can group different usages of "dog" based on sets of nearby words in the input data, then treat them as different inputs.

[Reelya]

https://news.slashdot.org/story/20/02/12/1628214/this-app-automatically-cancels-and-sues-robocallers

DoNotPay Founder and CEO Joshua Browder's Robo Revenge app is unique from every other app looking to protect you from robocalls in that it can get you cash while stopping them completely. "All of the big companies like AT&T and Apple have failed to protect consumers," Browder told Motherboard over the phone. "Consumers have to protect themselves. The only way the problem will end is if the robocallers start losing money every time they call someone."

In the past, DoNot Pay has offered various apps to help consumers fight back. DoNotPay's Free Trial Card creates a virtual, one-time-use credit card to protect you from getting charged by "industrialized scams" like free trials. DoNotPay's original offering was a chatbot lawyer program that automatically disputed parking tickets in small claims court. Robo Revenge combines both features to automatically add you to the Do Not Call Registry, generate a virtual DoNotPay burner credit card to provide scammers when they illegally call you anyways, use the transaction information to get the scammer's contact information, then walk you through how to sue them for as much as $3,000 per call under the Telephone Consumer Protection Act (TCPA), a law already on the books meant to protect consumers from calls that violate the Do Not Call Registry. The app also streamlines the litigation paperwork by automatically generating demand letters and court filing documents.

[Telgin]

That's pretty great.  I wonder if something similar applies for unsolicited text messages, since I've been getting a lot of spam from those lately, and the messaging app on my Android phone appears to have no way to block text messages from people not on my contact list.

It would probably be easier to just find a new app that did that.