re: Musk promise

quote:

---

And yes it does something on its own, it takes input and transforms it. That's what an algorithm is.

---

An algorithm itself isn't code, that's all I said the entire time, which you agreed with

Then you said it gets converted to code, which it can be and certainly does, that code is not the algorithm. But the algorithm remains, and it useless until taken and implemented into a larger set of code.

quote:

---

And yes it does something on its own, it takes input and transforms it. That's what an algorithm is.

---

Sure it can, it implies what type of code to write, but not, it does nothing.

And it would *do* something, it would give people an idea of how the application works.

As I've said, that would be disappointing. Not because the algorithm doesn't do anything (it does, it generates scores and visibility sorts), but because you're missing the most important parts.

When you tweet something, there are likely a bunch of classifiers run against it. They are built on models from hand-labeled classifiers. At some point someone probably had to say "this is abusive" or "this is racist" in a backend for certain tweets.

THAT is what would be interesting. The training data.

So you guys are going to continue to agree while not realizing that you just disagree on semantics for another 4 pages or are you going to let it go at some point?

quote:

---

You do, but you do not have the algorithm written as itself, only implications it was used to write the source code.

---

Most software algorithms are written directly in code. There are lots of well-known mathematical ones from days of lore that were written on paper. These are tried-and-true, like shortest distance algorithms

But 99% of Twitter's source was not put on a whiteboard first.

quote:

---

If they have any form of source code tracking like SourceSafe, it should be easy to see what it used to be and who changed it.

And any major gaps in the history record should be major red flags.

---

Thanks. Was just about to post the same. Plus, any changes would be traceable back to the person making the changes.

quote:

---

Whats going on behind the scenes right this instance to hide the current?

---

being open source has it's problems as it's more hackable, but if they are doing any manipulation on shadowbanning, removing followers, taking people out of mentions, or moving certain stories or hashtags to the top over more shared stories, etc. you'd be able to see if there was human manipulation to make that happen or algorithms in place.

e.g.

anytime somebody has said #XXX is trending but it's not showing as such or something that is obviously not trending (#thanksbiden) is probably because they have a walled garden approach to their code and have manipulated what shows up. open source code would let people see if something was manipulated to make that occur.


ETA: i see this has been more fully covered by more capable posters since i started typing this post. ignore. i'm a network guy, not a coder.

quote:

---

Plus, any changes would be traceable back to the person making the changes.

---

I'm not an IT guy, and I didn't even stay at a Holiday Inn Select last night, but is everyone dismissing outright the possibility that if the entire operation is compartmentalized, then it isn't impossible that the original algorithm parameters could be inserted from another "department" and that "department" is the bad actors?