Hacker News Reader

I don't know how well this would hold up in arbitration, but he has repeatedly cited the existence of bots as a reason for buying and fixing twitter.

> If our twitter bid succeeds, we will defeat the spam bots or die trying!

April 21 900K likes

https://twitter.com/elonmusk/status/1517215066550116354

4% of users could be very vocal and annoying bots...

Musk can take care of %4 but if its at %6 free speech is not salvageable?

For the right price maybe, for the wrong price no.

I remember a few weeks ago when the deal was first announced. A bunch of folks reported that conservative accounts were gaining followers while liberal accounts were loosing followers. The analysis back then is this could not have simply been organic because of when the followers picked up and dropped off, basically looked like a script doing the work.

This led to the theory that Twitter might have way more fake accounts then it's leading us on to believe. There was some speculation that the "Less than 5% figure" would come to bite them in due diligence so they were panicking and dumping bot accounts. Now this...

I'm not saying these two events are related, but there does seem to be something fishy going on. My gut says it will come out in the next few days that Twitter has something like %10-15 of their accounts being bots rather than the initial "Less than 5%" figure.

>The analysis back then is this could not have simply been organic because of when the followers picked up and dropped off, basically looked like a script doing the work.

Is there a link to this analysis?

Imagine if the old law of 1% applied here. That is on the internet, 99% of all content is generated by 1% of the users.

Now if we include "users" as "content" (bots, ai, scripts), we have a very brooding fringe theory similar to "dead internet theory."

In my experience... it's pretty obvious and prolific how much automation is in Twitter. It's basically a public cyber-war; you have state actors and non state-actors using military grade propaganda tools for 'reality framing' between military interests, government interests, and corporate interests.

Interestingly after just digging into this a bit more it seems that when considering this principle in multiple studies measured since the 2000s, the most successful predictor for this power law is if the internet property is a form of community. Specifically in those cases there was a strong 1-9-90% rule in effect.

1% of the users were responsible for all generated content in terms of creation. 9% were responsible for editing if they had the privilege to edit. Perhaps “replying” would fit in this segment. 90% lurkers.

This is a stronger version of the Pareto principle. It’s interesting metaphysically and perhaps physically when considering musical theory, intervals, and how boundary conditions seem to work pretty consistently despite seemingly unrelated things.

I digress though, back to my earlier point. It seems if these laws hold up then there is a strong potential that much of the users who are active are in fact “avatars of creation.”

It’s certainly been mentioned that the internet has felt like this over the recent years by more fringe echo chambers. However given the backdrop of progress, it certainly gives more credit to the idea of a more nuanced form of the dead internet theory reminiscent of “society of the spectacle.” What is the said progress I allude to? Well, “NLP” of course. The most amount of progress in machine learning and AI is specifically within language models. Given that the availability of building such automation tools to conduct memetic warfare doesn’t even require sophisticated language models, (manpower, shills, some creative spam) one may have to wonder for how long oneself may have been talking to themselves on the internet? In a way it forms an “imaginary great firewall” for every user. Interacting with avatars built on your predictors, with seemingly no truth vector for the utility function and it’s behavior modification goal (alignment functions). Weird. I think we are past the smoke alarm stage. The evidence of ability and active training is self-evident since 2016 on 4chan and elsewhere.

> this could be a relatively cheap way of mortally wounding Twitter

Interesting idea, albeit very high risk, however what's the motivation? Say he's right, sure he walks away with $1B minus his legal fees. But if he looses, he own Twitter... but wanted to "mortally wound" it.

Not sure he needs to do something high risk like this for the potential of $1B, and based on how much he uses Twitter I'm not sure he wants to take it down.

To be clear, I'm not even suggesting this is what Musk is doing. It's just an interesting thought I had about what someone could do.

But as for motivation for someone doing this, people have egos and when they have this much money, it essentially costs them nothing to exact revenge for no other reason than a bruised ego. Musk in parituclar is notoriously thin-skinned.

Remember when those boys in Thailand were being rescued from a cave following a flooding? Musk offered a submarine. One of the rescuers said something dismissive about it. Musk responded by calling him a "pedo guy" and then won the defemation suit by lying and saying it was South African slang, not an unfounded allegation the guy was a pedophile [1].

[1]: https://www.bbc.com/news/world-us-canada-50695593

Where does the article support your claim that he was lying?

That article doesn't.

However,

https://theoutline.com/post/7951/elon-musk-pedo-guy-south-af...

https://www.businessinsider.co.za/elon-musk-pedo-guy-really-...

https://slate.com/technology/2019/12/elon-musk-trial-pedo-gu...

https://www.iol.co.za/saturday-star/news/elon-musks-pedo-guy...

Worth noting that, even if it was common 'slang', the meaning hasn't changed one bit from what we attribute to those words today.

I was always under the impression that Elon was making a (poor?) joke about the sex tourism stereotype of bald Britons living in and around Thailand specifically.

Not every cunning idea has to be thought up in advance, and not every aspect of it needs to be devised by one person.

Perhaps some of Musk’s advisors dreamed up this strategy on realizing the dangers of “plan A” or even after the decision to back out was made.

>But if Twitter is found to be misrepresenting those numbers, they have way more serious problems than if the Musk deal falls through. They've then opened themselves up to litigation by the SEC and investors that they materially misled investors.

Well, yeah, this goes for every single public company (including Tesla).

I am curious how they prove either of the arguments about fake accounts precisely. What is the definition of fake account and how to prove them? If they can prove some self-evident cases, it could be just the tip of the iceberg.

It seems like there are 4 categories of Twitter users:

1. Blue checks

2. People who are “obviously” people (PII listed)

3. “Anonymous people”. I see a lot of theses, people with the name “iluvcrypt0” and an emoji of a Pokémon.

4. “Obvious” bots.

From looking at thread responses on twitter I think group 3 is probably the biggest, On the Luna thread someone posted to a link of posts that were implying they lost all their money due to Luna crashing, and it was a lot of group 3 posters. but of course each of these categories has a certain percentage likelihood that they are a bot. Im sure there are a few blue checks that slipped through and there are probably a few “obvious bots” who manually post.

The question is really the makeup of group 3. How many people who seem like anonymous users are actually bots, and can they prove that (when of course Twitter is incentivized to err low)

Also, either way, group 3 doesnt seem like they are contributing to the marketplace of ideas, they seem more like trolls/shit posters.

What about people who were banned and called Russian bots for their political opinions — Do they count as real people or bots?

> What about _people_

I think you answered your own question, if you’re using the word “people” then they’re not bots.

Category 2 if they used PII, category 3 if they didn’t.

An account that is shared between bots and humans is likely. More of a "bot augmentation". Eg a script that runs during certain hours or to handle high volume during certain periods.

Then it's almost a question of "how botty" the account in question is.

If this is true, where’s the proof of a conspiracy? This is similar to the thing about phones listening to conversations and then some app shows a relevant ad: where’s the conspiracy?

What you outline above implies something like an affiliate/recruiting program to find accounts to “co-house” your bot within. Especially with the implication that they operate on behalf of USian political parties.

I think we're probably using the word "bots" to include "human-based mass creation and use of twitter accounts". Like if I make 100 twitter accounts to promote <thing>, they count as "bots" here, even if I did it all by hand. If we're not, we should be.

Given that, I'm not sure there is much "bot augmentation" on otherwise human twitter accounts. It would be strange to do that if you were only using one account.

I think there's another category:

5. Users with plausible names and a profile photo of a person that are bots

I don't remember the study but when there were studies coming out of believed Russian-controlled accounts supposedly amplifying misinformation, all the examples I saw were in category 5.

And this category is the most deceptive. Nobody cares what a Twitter egg says. People care about supposedly grassroots outrage / support / etc. from supposedly legitimate people.

Yeah, I wasn't sure whether to include this as a false positive in category 2 or it's own category but yes your point stands.

I think there are many more categories. I am on Twitter with a pseudonym, and while it is not bot-looking, it is obviously not a real name. At first glance I may or may not be a bot run by somebody else. I can be Googled easily and you will find some PII there, such as my employer, but it would take some checking that you can't scale.

I am definitely between 2 and 3 in the above classification, and so are many of my contacts I would say.

[edit: can't* scale]

Also, if they already know there is X amount of bots, why don't they just...ummm...delete those accounts already?

There's no reason to go through every single account to find the true number of bots. At the scale of Twitter they could just randomly sample a set of accounts, run detailed analysis on them, and statistically extrapolate onto the entire userbase with confidence bounds.

I don’t think you could easily prove Twitter is materially misrepresenting the number of bot/spam accounts. Presumably, it is just an estimate based on some combination of assumptions and statistical analysis. You might be able to create a significantly higher estimate, but that seems different than proving material misrepresentation.

It's rampant tin-foil hat speculation, but the ordering of events may not be what we've seen.

Yesterday the big story was that two higher-ups in Twitter were fired unexpectedly, one of whom was on paternity leave at the time. Some amount of shake-up is normal in these kind of conditions, but that doesn't rule out something more. Today Elon's pushing back, suggesting the 5% bot rate may be inaccurately low. It's possible the events are related, some misrepresentation was found while investigating the figures, and that's why the firings happened yesterday.

If that happened, it'd be in Elon's interest to draw attention to it and that the deal is "on hold," as Twitter will be on the hook for that $1 billion breakup fee unless they can renegotiate terms favorable to Elon. And even if the deal dropped completely, now Twitter implicitly will be on the hook for misrepresenting to existing investors the percentage of bots for however long they've been doing it.

I guess we'll see. The truth will probably be more boring.

> If you think about it, this could be a relatively cheap way of mortally wounding Twitter. Make a buyout offer, get access to the books, prove they're lying about DAU/MAU, walk away with no termination fee paid, watch the executive team get sued into oblivion and the company tanks

That's true, but what does Elon gain from destroying Twitter?

I'd argue that Twitter actually was (and still is) a big part in building his cult of personality. Sure, it has its problems, but I honestly believe that he thinks he can improve Twitter.

Bots probably seem like they make a bigger part of Twitter's population b/c they will simply tweet more than the median user. I imagine that the majority, and perhaps the vast majority of Twitter users are mostly passive users, ie: they mostly consume tweets and seldom actually tweet themselves. In other words, high D/MAU, ~5% bots, but bot tweet volume makes up 30+% of actual tweet volume.

1 billion to just to say "nah I actually don't want to do this". 1 billion for, after all is said and done, nothing to actually materially change. Does that sound supremely fucked up to anyone else? All that fucking money...and it just goes to whomever bank is financing this because nothing happened?

What a massively fucked world we live in.

Absolutely not. Imagine I want to buy your house for, let's say, 200k$ and sign the contract. So need to quickly find a replacement, move out, clear the legal stuff and probably miss work. Then, at the last moment, I back out. Don't you think 5k$ would be appropriate for all the now useless work and money spent on your side?

It's the same number with Twitter, just scaled up to the actual offer.

I mean a lot has happened because of it. It's been a top news story across fields. It's affected Tesla stock. It's affected public perception of Musk. It has caused deeper analysis of the financial viability of Twitter. It has allowed new parties to get closer looks at Twitter's books. It may have left to the ousting of several Twitter execs. It has normalized a return of Trump to the medium.

I interpreted this comment as not saying a $1 billion penalty is necessarily inappropriate for this particular deal, but it’s a waste of money in general to pay $1B to achieve nothing when that money could have been put to better use or donated to charity.