Speech recognition, device miniaturization, and intelligence collection | Hacker News Reader

Speech recognition, device miniaturization, and intelligence collection

defenseone.com

56 pointseplanit25 comments

It will be interesting when all of a future political candidate's past phone calls, emails and chats can be pulled up and searched for anything and everything.

I've thought for a long time that in the future only the most boring people who have ever lived will be able to get elected to any high office. We're talking boy or girl scouts with perfect attendance, a 4.0 GPA, never tried any sort of drug, never curse, had totally dull sex lives, and have only the most mainstream beliefs imaginable.

I think it's more likely that the bar for "scandal" will be significantly raised. Consider what has happened in just 20 years where we went from Clinton who "didn't inhale" (marijuana) to George Bush, whose former use of cocaine was reported but never admitted, to Barack Obama, who admitted to marijuana and drug use in his own autobiography. I imagine that politician boringness and attitudes towards what is scandalous will meet somewhere in the middle, so you'll have slightly more boring people in office, but no so boring that people would care if they found out about some trips to a sex club or something.

No, who will be more likely elected to office are likely corruptible or already corrupt individuals whose vices are known levers for allowing their wealthy/corporate masters to control them easily and cheaply. Anything that breaks through can be controlled by a thoroughly corrupted corporate media.

This is the natural outcome when money is equated with speech... only the money has any say anymore.

> I've thought for a long time that in the future only the most boring people who have ever lived will be able to get elected to any high office. We're talking boy or girl scouts with perfect attendance, a 4.0 GPA, never tried any sort of drug, never curse, had totally dull sex lives, and have only the most mainstream beliefs imaginable.

It's actually worse than that because nobody lives that life. So what happens instead is that everyone has to pretend to but only the ones the people with the big database don't like have their deviation from the norm sent to the major media outlets for publication.

s/boy or girl scouts/puppets raised in a bubble by sociopaths/

... with policies dictated by the folks whose sociopaths are best at hiding behind their puppets.

It's going to be very, very ugly.

We'll have a lot more Eliot Spitzer's no doubt. Piss off the wrong person and suddenly you get singled out for using prostitutes. Even if the same prostitutes that are known to be popular with lots of other politicians. Only one gets annihilated.

What makes you think that hasn't already happened? The NSA recorded Washington area codes during the 2008 election campaign, ostensibly due to a typo:

http://www.washingtonpost.com/world/national-security/nsa-br...

That's already happening in some countries, with the exception that those countries don't have "political candidates" because they are frank one-party dictatorships. I would hate to be an organizer of a new civil-society organization in north Korea, for example, or even in China, which is still a dictatorship even though it is recently a less stark dictatorship than its client state north Korea.

Is it not interesting that Mr. Obama evolved from "Transparency promotes accountability and provides information for citizens about what their government is doing," to "The administration’s war on leaks and other efforts to control information are the most aggressive I’ve seen since the Nixon administration" [1]?

What could have caused this?

[0] http://www.whitehouse.gov/the_press_office/TransparencyandOp... [1] http://cpj.org/reports/2013/10/obama-and-the-press-us-leaks-...

It's pretty easy to predict what will happen. It will be a continuation of the current trend against diversity of views in US politics. The system will stagnate because security has been defined as the perpetuation of the system, not agility or adaptation. All nails sticking up will be pounded flat, using omniscient surveillance.

The NSA had a file/active series of wiretaps on Obama from at least 2004 onward.[0]

We live in interesting times.

[0]:http://www.businessinsider.com/the-nsa-spied-on-barack-obama...

I don't find that allegation credible. Russ Tice became a whistleblower in 2005, but never mentioned the future president was specifically and illegally targeted until boasting about it on radio program in 2013?

Other information produced by Tice at the same time was turned into a well-researched article that ended up winning the Pulitzer: http://www.pulitzer.org/archives/7040

It doesn't get much more credible...

He was a great source for an excellent piece of reporting in 2005.

What happens? The real targets stop using normal language, and start communicating in code. Replace "guns" with (say) "flour". Replace "ammo" with "fruits". etc. etc.

And the innocent folks' dirty laundry will be archived for future retrieval.

The article talks about hidden markov modelling, which combined with initial training data, can theoretically be applied to your cipher to solve for a solution. The number of things that humans can observe and create sensible (memorable) code from is not infinite. It must be based on some aspect of the environment from which information is derived. You get the abstract form (the grammar) and train to infer the patterns of how substituted words are selected.

Then, what will happen is language will cease to make sense as the complexity to communicate secretly increases. The real targets will communicate using literal forms of encryption. There will be layers and hierarchies of metaphors and symbolism, organizations of analogies and relations, similar to the complexity level of fictional narrative literature. It is a combination of abstraction and environmental observation.

Or maybe it is already happening, and it's just a little check mark next to a bunch of check marks that indicate something symbolic and abstract about the future.

It does depend on the size of the corpus; the literal interpretation could still be the highest-probability interpretation even if it's wrong. So if I have a habit of always using the word "norble" instead of "computer", someone with a large corpus of my speech can manually or automatically infer what I'm talking about (particularly given a knowledge of English, since I'd say stuff like "norble programming" or "reboot your norble").

If I speak an unknown language or jargon or relex or word-substitution code for only a short time -- giving very few samples of each unfamiliar word -- that might not be possible. There's just not enough data to make the correct interpretation more probable than other interpretations!

But yeah, if linguists and philologists can translate languages just from fieldwork observations or inscription corpora, we can be confident that if you use language in a weird idiosyncratic way for a while, people (or machines) who can observe you will come to understand you too.

A few years ago, I was reading about terrorists in Kashmir who realized they were being monitored, so switched to using common words for their terroristic communications. The same person, talking to X will use "tea bags" for "ammo"; and talking to Y, will use "tea bags", for, well, "tea bags".

Now he makes 2 calls: "I'm out of teabags; can you send some more?". How can you tell what he's talking about?

I would monitor and perform analysis for a grammatically structured pattern across the dialogue, instead of focusing on the individual person as a discrete entity. I would treat a dialogue from t(a) to t(b) as a unit to analyze instead, and look for markers that indicate a dialectically independent symmetry between the composition of sentences.

If people use word substitution, they are relying on an implicit pattern they can both distinguish "on the face" as code.

Time was, on the Internet, no one knew you were a dog.

Context: http://en.wikipedia.org/wiki/On_the_Internet,_nobody_knows_y...

I thought this article might talk about social, political, or ethical consequences of a world where spies can eavesdrop on any conversation, but it looks like a case where the editor gave the article an interesting article that doesn't quite match the content.

The article is mostly about how the government is continuing to invest a lot in speech recognition technology, how that investment is motivatived by the desire to have better full-text search of audio recordings, and how government researchers are excited by the possibility of using MEMS and/or IoT to spy on "populations".

So, not much of anything about "what happens when" they get those capabilities.

Can someone please suggest a more accurate title?

Hmmm, I wasn't even thinking about what good HN link title practice is. I remembered the idea that people should ideally use the same title as the linked article, but not the idea of trying to improve on it in the situation where the existing title is somewhat linkbaity or inaccurate.

Here, I think the summary you ended up with is an improvement!

Reply on news.ycombinator.com