Internet & Search
How to Easily Stop Internet Piracy: Make People Pay for Bandwidth
In the last days there have been so many debates and threads about the PIPA/SOPA legislation. Sites went black (even Wikipedia), people added “STOP SOPA” banner on their avatars, and the technology news is overwhelmed with those messages. There are tons of people who say that this legislation will destroy the web, freedom of speech, and innovation.
While I do not have a particular position on the topic, I do think that there could be a much easier way for the government to stop Internet piracy, which does not require any form of censorship: obligate service provider to make people pay for bandwidth instead of a monthly flat rate, pretty much like they are already doing with cellphone data.
Let’s assume that normal Internet users use 50 Gb/month and currently pay $50/month for the service. That’s $1/Gb of bandwidth. If you do a lot of file sharing, you will definitely use more, and you’ll pay for it. After a while people will stop because it will become too expensive.
In theory, this should work just fine. One could argue that I could legally download/watch things on regular channels (e.g., Hulu, Amazon, …) and those will still be counted towards my bandwidth effectively increasing my bill. True, but I think we could find a fair amount of traffic which covers pretty much everybody. My hunch is that people who really contribute to Internet Piracy use way more bandwidth than regular people.
Another solution could be to have a list of sites which do not count toward the bandwidth. Unfortunately, this would be hard to regulate and check, while I can always look at my router’s traffic stats to check that my cable bill is right.
Finally, there could be a legislation that requires cable provider to report to the authorities statistics on the traffic of people who go above a certain monthly bandwidth. I am sure the entire process could be automated, including building up a system that automatically estimates the probability that the user is doing something illegal.
Peer-to-Peer Online Jukebox with Songs from each User
Today we have Pandora, Spotify, GrooveShark, and a dozen other sites to listen to music online. Each has its own caveats, does not really let you hear the music you want, has ADs or a subscription model, …
I often wonder why nobody created an online peer-to-peer Jukebox where each user contributes with some songs. The songs will be played/streamed directly from the each other’s computers, so if only one user has a particular song, who wants to listen to it will have to wait its turn in a queue.
Assuming that all the songs in the system are regularly purchased, the software/system could be designed in such a way that the songs are directly played from the owner’s computer to the speaker of the listener. In this way, there would be no copyright infringement.
Song recommendations, suggestions, and personalized stations will be easy to implement on top of this, as well as creating hardware radios which directly play music from the service. The system could even connect to the various cloud storage system (e.g., iMusic, Amazon, …) and pick up your songs from there so you don’t need your machine on all the time.
Photos in the Cloud: Geolocation, Auto-Tagging and Virtual Albums
Yesterday I was thinking to my next photocamera. I want it small but with some buttons (no clumsy touchscreen), capable of taking pictures at night, waterproof, with integrated GPS, and capable of backing up the pictures automatically when I get home. Unfortunately, I do not think that it exists yet.
Then I started thinking about how I would like the computer to help me organize my photo collection. It would be cool if tags were automatically added to the pictures, for example:
- using the GPS location it could add tags like “Europe, Italy, Tuscany, Pisa“, and maybe also things like “beach” or “hill” perhaps also using the camera settings
- using the timestamp it could add “Summer, August, 11th, 2011, evening“
- using face recognition it could add “people, kids, crowd” or “head shot“, the distance of those faces (e.g., “closeup”, “far”, etc) and perhaps the names of the people in the pictures
Once the pictures are mostly tagged and automatically uploaded online, it would be great if I could create virtual albums through searches on the tags. Right now, all the online photo services that I know obligate you to upload one copy of the picture for each album. That’s insane! I just want to type a few keywords, give a name to the album, and share it with my friends, not upload the same pictures a dozen times.
For example, if decide to create an album to show my family my great summer in Europe I could just search for “summer 2011 Europe” and it will bring up all these pictures. I can “hide” the pictures I don’t like, add a name to the album and email the link. Done. If I want to create a romantic time-line of all the beach shots with my girlfriend I could just search for our names and the keyword “beach”, sort them by timestamp, and publish the link. Done. And if I need an head shot for my new online profile, I can just search for my name and “head shot”, create an album, and ask my friends to vote on the best picture for me. Done.
How long will it take before all this happens?
Facebook should Show Users what is Tagged before Publishing it
I think it is time for Facebook to get privacy right. The thing is that I do not think it is that difficult. A simple change could make almost everyone happy because they will be in control of what is published about them.
When a user is tagged in a video, picture or post, Facebook should not publish the tag immediately but send the user a notification. At that point the user should have the option to: (1) make the tag public, (2) make the tag private or (3) remove the tag.
Making a tag private is an option that does not exist at the moment, but that I think it would be extremely welcomed. It would allow users to personally keep track of all the posts, videos and images in which they have been tagged, without them showing up on their profile (e.g., maybe because it’s a bad picture but still wants to keep it around).
Why is it taking so long?
Forget the Filter Bubble, Personalization done Right will just Hide what you Really Don’t Like
In the last month more and more of my friends are talking about the negative effects of personalization for news websites and search results. Eli Pariser even wrote a book about it and sold lots of copies.
The points made in “The Filter Bubble: What The Internet Is Hiding From You” are correct but certainly not new.
Most of the personalization systems today are based on collaborative filtering, a technique made famous in 2007 thanks to a paper by Google. This approach is based on the assumption that somewhere in the Internet there are some other users similar to you, and what they like may interest you as well.
For example, if John likes A, B and C, and Bob likes B, C and D, it’s very possible that John may find D interesting and that Bob may like A as well. Pretty simple, right?
The central argument made by the book is that these systems are recommending to people more and more of “their kind of stuff”, which makes it hard for them to reach/discover something outside of their comfort zone. Touche. If things are not done correctly, that is exactly what will happen.
Personalization done right, should actually do the opposite: hide things that you will definitely not like
A personalization system designed correctly will hide only the things that is sure the user will not like, and there is really no harm in doing that. What is left, needs to be ranked appropriately. It is ok to rank higher what the system believes the user will be interested in, but it is necessary to keep throwing in a few outliers, to monitor the user response (perhaps she clicking on everything) and avoid to end up in local optimums.
It does not just apply to news or search results, but to every recommendation/personalization system. That house in Seattle on sale for $3 Million? Out of my price range and in a place I don’t like, so, hide it. That book about knitting center pieces? Not likely to interest me right now, so, ditch it. That news about New Zeland election of a new Prime Minister? I can probably wait to hear about it, so, screen it out.
Hiding things that I do not like will help me discover more of the stuff I might like. In addition I am sure to do not miss out on things that I definitively am interested in.
Creating such a system is clearly a bit more complicated than building a vanilla collaborative filtering algorithm, so, make sure you do not go cheap while picking your Artificial Intelligence guy.
