3 ms·

Amazingly Naive. Just the data and code needed to make the “algo” scale to what it is means there is no good way to “open it up”. But let’s say, and why not,

Amazingly Naive. Just the data and code needed to make the “algo” scale to what it is means there is no good way to “open it up”. But let’s say, and why not, that it actually was opened. I can see it now, everyone and their mother would be recommending changes to it. People would want it at the extremes, or tweaked to just not recommend their pet peeves. And if they did not get their way, they would go off in a huff and threaten to join another service. Years ago I ran a big enough service that had millions and millions of users every month. A member of the military came on and demanded we allow him to do something that was against TOS. He complained that this was what he fought in Iraq for. So he could write anything he wanted anywhere. America is certainly not free and it’s time we stop giving our content to services like Twitter and Facebook if we believe that.

extheat 4y ago
> I can see it now, everyone and their mother would be recommending changes to it. People would want it at the extremes, or tweaked to just not recommend their pet peeves. I think the solution to this is to let people switch between different recommendation algorithms, a stream of chronological tweets, most liked among recent posted, one powered by ML, etc. I don't see any reason the implementations behind this, or even Twitter more broadly, could not be open source. There are many other open source projects, like Signal that manage this just fine. And for a while, even Reddit used to be open source. It's definitely do-able.
- curioushacking 4y ago
  Exactly. While ML interpretability may be tough, I don't really read it like that. More here are X number of algorithms and each one is aiming to do Y. Chronological, Engagement, Happiness, Current Trends, etc.
bko 4y ago
I think they're looking for a statement like: if (tweet.from in bad_users): score *= 0.8 Do they have the levers in place to manipulate results or is it a clear objective scoring system that shows what you see on your feed and search results? I believe they have something like this. To what extent its being manipulated is a separate question.
- alm1 4y ago
  that's why a lot of responders are calling the approach naive. This function is potentially the last layer in thousands of functions. Thinking that one can predictably change the behavior of a large nonlinear model via one variable is like thinking you can change the US political system by passing one good law in Reno, Nevada. This approach only works if auditability is mandated to also be a property of the model, which typically reduces model complexity and accuracy too. We do this with credit scoring models semi-successfully.

extheat 4y ago

> I can see it now, everyone and their mother would be recommending changes to it. People would want it at the extremes, or tweaked to just not recommend their pet peeves. I think the solution to this is to let people switch between different recommendation algorithms, a stream of chronological tweets, most liked among recent posted, one powered by ML, etc. I don't see any reason the implementations behind this, or even Twitter more broadly, could not be open source. There are many other open source projects, like Signal that manage this just fine. And for a while, even Reddit used to be open source. It's definitely do-able.

curioushacking 4y ago
Exactly. While ML interpretability may be tough, I don't really read it like that. More here are X number of algorithms and each one is aiming to do Y. Chronological, Engagement, Happiness, Current Trends, etc.

bko 4y ago

I think they're looking for a statement like: if (tweet.from in bad_users): score *= 0.8 Do they have the levers in place to manipulate results or is it a clear objective scoring system that shows what you see on your feed and search results? I believe they have something like this. To what extent its being manipulated is a separate question.

alm1 4y ago
that's why a lot of responders are calling the approach naive. This function is potentially the last layer in thousands of functions. Thinking that one can predictably change the behavior of a large nonlinear model via one variable is like thinking you can change the US political system by passing one good law in Reno, Nevada. This approach only works if auditability is mandated to also be a property of the model, which typically reduces model complexity and accuracy too. We do this with credit scoring models semi-successfully.