What's 2 + 2? Personal info, sniffs Twitter: Anti-doxxing AI goes off the rails, bans tweets with numbers in them

Netizens are being locked out of their Twitter accounts for tweeting innocuous posts and images, such as math equations, that trigger the social network's system that prevents the sharing of private personal information. Revealing personally identifiable non-public information, such as someone’s home address or cellphone …

[Sorry that handle is already taken.]

Simplify all the things

What’s common with all the offending images is that they involve a white box against a dark background. “The key is a centered white box, and a dark color around the box. Then your account will be locked,” AltentX added.

Isn't a white box against a dark background the "spherical horse in a vacuum" of machine learning?

[stiine]

Re: Simplify all the things

No, but if you don't add an avatar to online accounts, most of them default to a silouette, a blank white head on a dark background. Twitter must have some wonderful DevOps people....

[Anonymous Coward]

Re: Simplify all the things

"Isn't a white box against a dark background the "spherical horse in a vacuum" of machine learning?"

I'm not the smartest person so I don't know, but I can say that from my whopping 2 "hello world" tutorials with OpenCV, I can't help to think I've taken 2 more tutorials than Twitter.

A 100% guess at your meaning of "spherical horse", but yes, you kind of want as few contrasting colors as possible to increase accuracy, or in Twitter's case... not to.

[DS999]

They will never automate this no matter how hard they try

It can do pattern matching for "this has already been found to be something we've banned" but the idea that an AI can be trained to determine what is or is not doxxing is ludicrous. If I tweet the address of a voting site, or a historic building that was destroyed in a storm, how can the AI tell I'm not doxxing whoever lives there? How is going to handle bad actors who resort to code words like "the drug store at 111 E Main St sells Tums" and have that mean that's where someone skinheads should beat up lives?

[Anonymous Coward]

Re: They will never automate this no matter how hard they try

It isn't. That doesn't matter though, that just shows we need evan moar of it¹, and possibly make posting pictures that do not get whiteboxed illegal. Don't you think of the chiiiiildren?

¹ it's the American Tourist approach to policy: if they don't understand, SPEAK LOUDER.

[codejunky]

Re: They will never automate this no matter how hard they try

@DS999

"the idea that an AI can be trained to determine what is or is not doxxing is ludicrous"

Very true but beyond the understanding of some people who are demanding facebook and twitter police the postings of billions of people, bots and lizard people. Unfortunately more censoring is demanded and thats in the west. As reg has posted, the moral police of other countries demand their own additional censorship- https://www.theregister.com/2020/09/01/pakistan_youtube_block/

[Lee D]

AI at its finest, then.

Completely devoid of understanding, meaning, context, but instead just choosing things based on arbitrary learned properties of the training images (e.g. black text on a white image).

There will come a day when people will realise the inherent weakness to all AI (zero inference), and that even the "AI" we have is basically just a large, dumb statistical model on unknown parameters that it "formed" itself, with zero insight into that choice, or any ability to modify. For all you know, it's detecting copyrighted images by the second-from-the-left top pixel being slightly green.

[General Purpose]

AI isn't inherently weak, just as generating power from nuclear fusion isn't impossible. After all, this is being written by a rather dumb device with limited powers of self-modification, formed by randomly merging codebases with little insight into the process, and ultimately dependent on fusion power. The one that's reading it may be more developed but not by very many orders of magnitude.

[Yet Another Anonymous coward]

When it comes to things like detecting abnormal cells in a smear test I'll take 'massive dumb statistics' over a bored technician giving it a quick glance

[jgarbo]

The Solution is Simple

Ban all numbers, symbols and alphabetic characters on Twitter. Only "mindful neural resonances" permitted. Now monetize that ...

[Imhotep]

Re: The Solution is Simple

We're are being driven to communicate using only emojis.

[find users who cut cat tail]

Re: The Solution is Simple

Fortunately, the nice people developing what3emojis have aleady thought of that.

I didn't know such thing existed before reading your post. But a few seconds after saying ‘now we just need a way to encode addresses using emojis’ I realised this is so silly that it must exist. And sure it does…

[gnasher729]

No AI

It should be clear to anyone by now that there is no such thing as Artificial Intelligence yet. We have Artificial Stupidity.

Really, for Artificial Intelligence to happen we need a lot, lot more. We need real _understanding_ and we are nowhere near that. We actually move further away, because the technology supporting Artificial Stupidity are progressing at an incredible rate, and all the computer science courses teach impressionable youngsters about Artificial Stupidity.

[USER100]

Re: No AI

> Really, for Artificial Intelligence to happen we need a lot, lot more. We need real _understanding_ and we are nowhere near that.

I would go further and say it can't be done. It's taking scientists a long time to reach this conclusion but I think they are getting there.

When computers first appeared, people could be forgiven for thinking that one day a powerful enough machine might be able to reason like us. As time has passed and computing power has grown however, it's become clear that that's not how the brain works at all. Somehow, biology is required to produce awareness, a phenomenon that cannot be replicated by algorithms. When Deep Blue beat Kasparov, it didn't think to itself 'Have that, Garry'. It didn't even 'know' it had won.

Of course AI is still massively important, but please don't ever expect it to actually 'know' or 'think' anything.

[jmch]

Re: No AI

Agree. No one really knows what consciousness is or how it emerges. Saying that it emerges from complexity is a random stab in the dark, a delusion that a computer wired the exact same way as a brain will behave the same. We are many many years from being able to successfully simulate brain hardware and we still are even further from understanding the software.

And even then there is no guarantee that simulating a human right down to the cellular level will generate consciousness

[Alan Brown]

Re: No AI

"And even then there is no guarantee that simulating a human right down to the cellular level will generate consciousness"

if you wish to argue that only humans display consciousness, then you're missing a lot. Advanced conciousness perhaps, but AIs still haven't gotten much past cockroaches anyway

[PapaPepe]

Re: No AI

Artificial intelligence does exist; just like artificial insemination. Both are substitutes, used by those not capable of the real version of the phenomenon.

[Man inna barrel]

Re: No AI

All that is required for AI to succeed is for people to trust computer programmes more than they trust their fellow humans, or their own common sense. There are still challenges here. We had some success with satnav, with the computer instructing someone to drive into a public toilet, or into the sea. But these are mere feasibility demonstrations. There should be a proper training programme, so that people stop thinking altogether. Significant progress is being made in education to this end, but we have to be patient.

[Man inna barrel]

Re: No AI

What we need is for people to think for themselves, and not blindly trust what the computer says. People doing what the satnav says and driving into the sea is tragic rather than comical. I suppose AI will become a reality when people become so dumb that computers can overtake them.

[Terry 6]

Re: No AI

it's the ones who drive off the road and into a field or past a warning sign saying "road flooded", because the sat nav says to go that way that gives me most concern about the future of the species.

[Mike 137]

"He reckons there might be a bug in the automated methods Twitter uses..."

Not arf!

Over a year ago I opened a twit account to promote our open source software, but it literally never got off the ground because:

[1] our user name was silently rewritten so it was meaningless

[2] after half a dozen attempts to get our avatar image displayed correctly (without success) the account was locked

[3] twit never responded to our complaints or enquiries, including a GDPR statutory request for erasure (article 17)

Consequently there's probably a potentially publicly visible garbage account still out there with some personal information in it. So much for twit protecting privacy.

[tiggity]

they do not care

Until financial penalties get applied for an account being erroneously silenced, teh likes of twatter, FB etc. have no incentive to have accurate AI algorithms. False silencing costs them very little so they do not care.

[Terry 6]

Context

The rapidly deteriorating usefulness of Google searches is pretty good evidence that computer based discrimination of context is not improving.

In general, what appears to be happening is a bigger and bigger dataset of what people asked for previously being used to direct results towards ignoring what the user puts in and instead feeding back what the user should have put in, according to their database of previous requests by millions of people who aren't that user and the need to give advertisers an appearance of value for money.

In The case of Twitter this seems to include an element of "Everyone is looking at or posting <this> this so when you ask for or post something different we'll assume it's just <this> anyway."

[Lon24]

Re: Context

Yep - back in the early days of Adsense we got a jolly good income from embedding ads on our websites taking advantage that the content was matched to ours. So the user aleady had an interest - and clicked.

Then they changed to the content 'matching' the user not the website so often producing stuff that was the antithesis of what the containing website was about. Our revenue slumped so much and as the contrasting content devalued our sites we exited the Adsense business.

The mystery is how this made Google richer.

[Anonymous Coward]

What's 2 + 2? Personal info, sniffs Twitter

Ahem. Quite right.

[Blackjack]

Would anything of value be lost if Twitter was gone?

I am honesty curious.

[codejunky]

Re: Would anything of value be lost if Twitter was gone?

@Blackjack

Since it is considered desirable by its many users the answer would be yes.

[Captain Obvious]

Re: Would anything of value be lost if Twitter was gone?

I have mixed feelings on this. I personally think Twitter/Facebook/Instagram/whatever are all stupid and a waste of time. Like others, I have tried to use it for promotion, but had too many issues.

The mixed feelings are: does it really help in an emergency and people are trapped in a collapsed building/cave/whatever or is it better to simply text or call if possible? If cell towers are out - none of these options are even available.

On the other hand, you can get rid of Trump's tweets - although maybe people look forward to their daily laughs... Again, I do not know.

[USER100]

Re: Would anything of value be lost if Twitter was gone?

I also have mixed feelings about it. While not on social media myself, I can see how it might be useful in situations like if somewhere was getting bombed or whatever, people could share that news without the filter of the mass media and its sponsors. Maybe we just have to take the bad with the good.

[Terry 6]

Re: Would anything of value be lost if Twitter was gone?

There was a brief window of time when companies who'd just about perfected ways of avoiding being told about customer complaints* could be contacted on Twitter and shamed into sorting out the problem they'd caused. But then it all got big enough for them to work out ways of bullshitting or ignoring that, too.

-------------------------------------------------------------------------------------------------------------------------------------------

*We all know this, probably suffered it too. But it needs to have the light shone on it.

Big companies don't want to pay customer service staff. So they hide or remove phone numbers and email addresses. Instead there's a web page with a tab that says "Contact us" that leads to a page of FAQs that have no relevance to anything that anyone would care about. followed, possible only after you've clicked on one of these, by a link that says "Need more help". This takes you to a generic Help page. On that page, carefully hidden, will be a contact us link. Which leads to the FAQ page.........

[Ian Mason]

Re: Would anything of value be lost if Twitter was gone?

> Since it is considered desirable by its many users the answer would be yes.

There used to be a style of TV advertising in the 60s that went "Use Zappo! Two million housewives can't be wrong!" - which I always characterise as the "Eat shit!" argument: "Eat shit! Twenty trillion flies can't be wrong".

So it would seem that your thesis for the usefulness of Twitter is "Use twitter! Three hundred million twits can't be wrong!".

[Mark192]

Re: Would anything of value be lost if Twitter was gone?

"Would anything of value be lost if Twitter was gone?"

Yeah, it's 99% junk like anything else but...

With breaking news, Twitter is a fast source of direct reporting from people at the scene.

When the Manchester Arena bombing happened, I got confirmation of it being a bomb from a brave young woman posting on Twitter when the BBC et al were still (deliberately) mis-reporting that it "may have been balloons popping" (you may remember the picture of her handbag).

When my kid's school is closed through snow, Twitter is the first thing to get updated.

Also:

When people have a complaint about service from a company, the people operating their social media accounts can resolve the issue when the zombies on the phones can't even understand the problem.

When a prominent person posts on Twitter you get it unfiltered, not as miss-reported through a partisan or sensationalist press.

But it's that first one that's important to me. When we were being deliberately misled, a single post by a young woman cut through all the lies.

[Jeffrey Nonken]

Tempted to get a Twitter account, tweet a white box with a black border, and see if it gets suspended.

[Anonymous Coward]

MAGA2020

now I know why my MAGA2020 got me banned. it was the darn 2020

[Mark192]

Two options

1. Fast and automatic but with a manual review (and continually updated so it gets better)

2. Too slow for the safety of the affected persons

Given that 2. will eventually result in serious harm or deaths, I can see why they've rolled out 1.