Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The accidental genius of Yo

capiche.com

Post reply on HN
Print view View original

71–80 of 94 posts

Re: The accidental genius of Yo

#71

Post by kybernetikos »

I noticed the gradual reduction of message length too when Yo came out (email -> tweet -> yo), and wondered what it would mean to transmit less than a single bit of information.

I came to the conclusion that it would be a messaging system, somewhat like yo where sometimes it would send a notification even when I hadn't told it to. If you configured it so it sent an extra notification randomly for every notification you sent, then you're conveying approximately half a bit of information.

Strangely enough, I think this could be quite useful, as a way of initiating conversations with people you would otherwise lose contact with.

Re: The accidental genius of Yo

#72

Post by SamReidHughes »

StavrosK wrote: post #67
SamReidHughes wrote: post #51

Earlier quoted context omitted.

That’s not one bit of information. If there is only one possible message you can send, there’s zero bits in the message itself. One bit would be if you had two messages to choose from. All the information in Yo is in the timing.

You do have two messages to choose from. One is no message, the other is a message.

In that case you have 1 bit per nanosecond of information, or however frequently you want to model sampling the message, not 1 bit of information.

I mean ultimately you need to somehow map the sequence of Yo messages into "information" in a fairly precise manner. Receiving the message takes you from one probability distribution about the world -- or some particular question about the world -- to a smaller probability distribution. I'd describe the information the message carries based on that. It's contingent on what the question is.

Re: The accidental genius of Yo

#73

Post by btrask »

tlarkworthy wrote: post #37

The article repeatedly makes the mistaken assertion that sending a 'Yo' constitutes 1 bit of information transmitted. This is mistaken as the information also holds temporal information, which is actually what is very useful. It's not a Yo at some random point, it's a Yo now. It's really encoding "now!". Still not very much information but a ton more than 1 bit, its a timestamps worth in fact... /pedant

Guys, I worked it out. It's the Heisenberg uncertainty principle/Fourier conjugate.

Let's say you send a yo now, and then you're going to send another one in 5 minutes. Then you decide to send a third in between. How much information does it contain? Less than a full timestamp's worth, because the timing is constrained.

Once you are sending yo's at the maximum line rate (certainly at least every nanosecond, it's such a critical piece of infrastructure), sending or not sending an individual yo is down to one bit of information. The time component is basically completely gone.

In other words, you cannot know the precise position of a yo at the same time as knowing the precise momentum.

Re: The accidental genius of Yo

#74

Post by stevenhuang »

thaumasiotes wrote: post #64
tsimionescu wrote: post #63

Earlier quoted context omitted.

You're arguing past each other. The amount of information transmitted is 1 bit. Someone receiving a PCAP with the data would not get more than 1 bit of information. However, you're also right that the people sending and receiving the bit can derive more information from it.

> You're arguing past each other. The amount of information transmitted is 1 bit. I don't understand the argument here. In order to get up to one bit of information, you need to take account of the fact that "yo" is only sent at particular times. If you ignore that fact, sending "yo" provides exactly zero bits of information, because it is the only possible message. All of the information content comes from the exist…

It's called a side channel.

The amount of information _sent_ really is 1 bit.

Re: The accidental genius of Yo

#77

Post by anonytrary »

sixstringtheory wrote: post #46
tlarkworthy wrote: post #37

The article repeatedly makes the mistaken assertion that sending a 'Yo' constitutes 1 bit of information transmitted. This is mistaken as the information also holds temporal information, which is actually what is very useful. It's not a Yo at some random point, it's a Yo now. It's really encoding "now!". Still not very much information but a ton more than 1 bit, its a timestamps worth in fact... /pedant

“now!” would be just a timestamp. Any additional message content adds context, so your assertion is wrong.

That was his point, right? A yo-train is just an array of increasing timestamps tied to the sender and the recipient:

  {
    to: 12,
    from: 15,
    messages: [
      1604885969953,
      1604885969964,
      1604885969973,
      ...
    ]
  }
The message is "factored out" and contains no information since it's the same every time.

Re: The accidental genius of Yo

#78

Post by thaumasiotes »

stevenhuang wrote: post #74
thaumasiotes wrote: post #64

Earlier quoted context omitted.

> You're arguing past each other. The amount of information transmitted is 1 bit. I don't understand the argument here. In order to get up to one bit of information, you need to take account of the fact that "yo" is only sent at particular times. If you ignore that fact, sending "yo" provides exactly zero bits of information, because it is the only possible message. All of the information content comes from the exist…

It's called a side channel. The amount of information _sent_ really is 1 bit.

> The amount of information _sent_ really is 1 bit.

I'm not sure how you're calculating this, but it is obviously wrong.

The information conveyed by the contents of a message is the log of the number of possible messages.

Here, there is one possible message. The log of one is zero. There's a big difference between one bit of information and zero.

You can interpret a "yo" message as conveying one bit of information, and the only way to do that is to draw it from the contrast between "sending a yo" and "not sending a yo". If you do that, "sending a yo" sends one bit of information, and "not sending a yo" also sends one bit of information.

Thought experiment: with three bits of information, you can disambiguate between eight possibilities. Try to design a protocol in which, by sending up to three "yo"s, you achieve 8 distinct messages.

Re: The accidental genius of Yo

#79

Post by maguay »

spiznnx wrote: post #62

This reminds me of feature I liked about Google's Duo app (like Facetime): send a heart. It's pretty clear what the heart means in the context of a video call app: "I am currently thinking about you and wouldn't mind a video call, if you want to." It solves the problem of instant-anxiety of getting an impromptu video call when you weren't expecting one. Unfortunately now you can send one of 10 different emoji instead…

That would be a great feature—random video calls are absolutely worse than random phone calls since the former requires you look presentable.

Re: The accidental genius of Yo

#80

Post by maguay »

kybernetikos wrote: post #71

I noticed the gradual reduction of message length too when Yo came out (email -> tweet -> yo), and wondered what it would mean to transmit less than a single bit of information. I came to the conclusion that it would be a messaging system, somewhat like yo where sometimes it would send a notification even when I hadn't told it to. If you configured it so it sent an extra notification randomly for every notification y…

Author here, agreed, and using something like that to stay lightly in touch with people ideally would have potential. Strangely I think it could have possibly worked in Facebook better if they hadn't named their "notify someone without a message" feature _Poke_.
Post reply on HN