Hacker News new | past | comments | ask | show | jobs | submit login
[flagged] U.S. Unleashes Military to Fight Fake News, Disinformation (bloomberg.com)
41 points by chadmeister 50 days ago | hide | past | web | favorite | 29 comments

The title is absurdly misleading. It sounds like this project is an effort to detect deep faked content, by DARPA. The article does a ton of editorializing around that.

Ironic, considering the topic of the article.

Titles like this drive clicks, as evidenced by the fact that this is now at the top of HN.

I don't like the title either, but the internet is built to incentivize behavior and content like this, unfortunately.

s/the internet/the economy/

IP definitely has its failings, but the incentive behind clickbait is a late stage economy answerable only to its own metrics.

On a technical level, "The Web" (HTTP/JS) is much more at fault than "The Internet", as its naive designs have fostered a surveillance industry that transmutes a negative-value pageview into positive money.

This was my favorite editorial-posing-as-neutral-news sentence:

> as the top Republican in Congress blocks efforts to protect the integrity of elections.

(translation: democrats want to protect our elections, republicans are russians)

I wonder if it will be able to detect massive journalistic malfeasance by people who perpetuated multi-year-long rumors about pee tapes, or whatever other politically motivated "news" posing as journalism

A literal statement of fact triggering that reaction says more about your filter bubble than the author of that article. Instead of just dismissing it you could talk about why don’t think the bills he’s blocking won’t work, how you don’t believe the reported threats are real, or the politics behind McConnell refusing to let legislation move forward.

Came in to post just this. I would not expect such sensationalist/incorrect headlines from Bloomberg, but here we are.

The Defense Advanced Research Projects Agency (DARPA) is a part of the military. If it is launching a project to fight dis/mis-information then isn’t that an accurate title?

It’s a research arm handing out grants. That’s very different from what the title implies, which is as if it were like any other war/operation with top active command and a ton of resources. The use of the word unleash implies there’s some strong existing momentum just being held back by an arbitrary gate.

DARPA funded the internet. Did the US unleash military to network computers? I think we could all agree that’s hyperbolic.

No, it's not accurate. It uses words aimed to give impression which is incorrect. "Unleashing military" means, in common parlance, physical attack of the armed forces, not fundamental research branch of DoD initiating a project. DARPA is a research division which initiated a lot of projects having wide non-military applications. Specifying that it is DARPA initiating the research project is a vital piece of context which changes the whole understanding of what is going on.

It's like publishing a headline saying "Elvis Presley sighted in Moscow", forgetting to specify that it's not the Elvis everybody thinks about, but some other guy that happens to have the same name, and has lived in Moscow, Arkansas, for his whole life and nothing of note happened to him ever. Classic clickbait.

This a a grant program running by DARPA called Semantic Forensics. [1]

> The Semantic Forensics (SemaFor) program will develop technologies to automatically detect, attribute, and characterize falsified multi-modal media assets (text, audio, image, video) to defend against large-scale, automated disinformation attacks.

The primary use case for this is election security, but more generally being able to detect and defend against an adversary trying to seed political discord with fake accounts, or fake news, so-called "deep fakes", etc.

[1] - https://www.grants.gov/web/grants/view-opportunity.html?oppI...

By "unleashing the military" they mean a DARPA research program.

Straight to the top of HN!

Tangentially related: what's the name for the approach where you analyze word choices and sentence structure in order to fingerprint a person?

Forensic linguistics? This technique was invented during the manhunt for the Unabomber. Features in the Netflix series of the same name.

Forensic linguistics or linguistic fingerprinting.


That's the one, thanks!

I am interested in the technical aspect of implementing such a system.

I know the founder of https://www.fakeskiller.com which is a browser extension for highlighting fake news (they also released a mobile app not so long ago).

They use fact checking organizations to classify news as fake news and highlight posts in social media, which contain links to fake news. They also provide link to rebuttal.

But since there is always a lag between producing clickbait fake news and working on proving it's a fake, I assume those who fight disinformation and propaganda will be on the losing side.

That's why the founder of Fakeskiller always invites everyone on educating those "useful idiots", who are sharing those fakes and sharing truth with them.

But can the propaganda be effectively fought technically? I assume, no.

Non-clickbait title: US DARPA finances a project that aims to detect disinformation campaigns.

I'm a bit confused on what sort of authorities anyone's got here or what the actual intent of the program is. Ultimately, does DOD have any sort of authorities to do anything about potential US person speech and how would that be constitutional? They have a need to understand who's doing what in various theaters, but I'm not sure why anyone would want DOD messing with content in the US.

They aren't necessarily going to (try to) defend against misinformation. You'll also want to study defense if your objective is offense - knowing how the other side will defend will hopefully lead to finding ways around that defense.

Emphasis on "fake news" and fact checking is so misguided.

The core problem is unsourced, unattributed information. The difference between gossip and journalism.

What is the provenance of every statement?

The Correct Answer is adding signatures to every thing, so we can independently verify that someone said something.

Once we have traceability, people can figure out what is more true (less wrong), over time. Just like with academic publishing.

Relevant video by Smarter Every Day[0]. Basically interviewing military people talk about the information part of war in the modern era.

> I just made a weapon?!

-Dustin @21:35

[0] https://www.youtube.com/watch?v=qOTYgcdNrXE

There should be a clear distinction between news that are factually false and those that share viewpoints that are counter to the mainstream. Labeling the latter "disinformation" is a mere act of political speech.

I respectfully disagree.

I just want two things.

1/ Show your work. Data, code, citations, attributed quotes, whatever.

2/ Sign your work.

Then everyone so inclined can determine truthiness at their leisure. Just like with academic publishing.

Omitting 1 means it's just opinion, 2 means gossip, both means propaganda.

There should be. We've tried it with "fact checkers". Most devolved into political operators rather quickly. Apparently it's not so easy to do, and probably not because we lack technology.

What could go wrong?

>to defend against large-scale, automated disinformation attacks.

I realize cable news has become entertainment without even trying to correct themselves the next day, but if it was really viable to automatically generate a quasi-ARG experience, wouldn’t Hollywood have developed it first?

I for one welcome the coming onslaught of fake news, at least I won’t have to question the motive of the news source.

Will they give the actual civilian toll after drone attacks on Yemen?


1984 is what’s actively happening in China right now, between constant surveillance and the large scale filtering of their internet.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact