Kevin Unhammer

The Apertium Project
Sandnes, Norway

About Kevin

Languages

Norwegian

Areas of Expertise

Computational Linguistics, natural language processing

Talk to me about

Free and Open Source Machine Translation

Favorite talks

Comments & conversations

90490
Kevin Unhammer
Posted about 3 years ago
David Birch: A new way to stop identity theft
"The initial set up (and subsequent running) would require proving to some institution every detail about your identity" Why? You could easily give some info to one institution, some other to another. It would just involve that the checker trusts that the info from that institution is reliable. So I give my age to some government-run repository, along with my public key. My name as well is stored there, but the government already has those facts in some database or other. Perhaps my employer stores my public key along with the "works here" fact. Still there's not more info on me in one single database than there already is; only now my public key is stored both places. Note that you can (and probably should) have several public keys, one per institution, so they can't be trivially linked. But logging usage records, that seems like a deeper concern. Like you say, no one but the two involved know that an age-verification happened when you show your drivers license. I can't think of a good way to counter that. (Other than to randomly run queries to the databases (from random IP addresses) so that "real" or "fake" queries are impossible to tell apart. But that would put more load on the system, require users to run anonymising services, and is not very elegant at all.) Hmm. Difficult. EDIT: I see the paper has a different technical setup, see the heading "Magic" on page 12. In the setup in the paper, the fact "over 18" (along with your picture) is to be stored in the actual SIM card (or some other physical, "tamper-resistant chip" ). I guess it would be fairly resistant to counterfeiting if the fact and picture were both digitally signed by some mutually trusted third party (the bouncer at the club just has to check the digital signature against his/her copy of the public key of that trusted authority who did the signing, and look at the picture transmitted from your phone and check it against your face. Seems to fix the logging concerns. Would love more info on that
90490
Kevin Unhammer
Posted about 3 years ago
David Birch: A new way to stop identity theft
Iit's not reliable. Faces can be changed with plastic surgery, and even change naturally, over time. Even people who are not mute can get colds or lose their voice in other ways. For automated systems, I can just take your picture and record your voice, and I can get past the door. (Some people log into their computers with a fingerprint reader, a singularly bad idea; why would you want a password that you've "written" all over your computer? To protect it, you'd have to wear gloves everywhere except when you log on.) And how would your face+voice prove that you're over 18? You'd still need a link to some trusted database, ie. a computer. If all that's needed is a face+voice, I could get at anyone's info just by posting their face+voice picture to the database, and get at _all_ the info, not just the fact that they're over 18. A real device (doesn't _have_ to be a phone in my opinion), however, can have your encrypted private key stored on it. Only you know how to unlock the key, so unlike your face/voice/fingerprint it's not accessible to anyone but you. And when you want to prove that you're over 18, you use public key cryptography (look it up on Wikipedia, in short: keys come in public/private pairs, the public key is used for encryption, and the encrypted text can only be decrypted by the private key belonging to that pair). One simple method: your public key is stored in the database along with the fact "over 18". You say you're the owner of that exact public key, so the database uses your public key to encrypt some random text, and hands you the encrypted text. If you are able to decrypt it with your private key, you must be the person whose public key was stored with the fact "over 18".
90490
Kevin Unhammer
Posted about 3 years ago
David Birch: A new way to stop identity theft
You're still you, a unique and complete human being. The point isn't about fragmenting your identity as a human being, but about not giving away more information than you want to. If you want to give away your real name, then go ahead, but you shouldn't be forced to just to buy a beer. I would not feel like I'm using a "hidden identity" or pretending to be someone else just because I prove to you that I'm over 18 without proving that my name is so-and-so. The "person who is over 18" and the "person whose name is so-and-so", etc., are still the same unique person, even if the bartender doesn't know that. And what makes the name so special? I could just as well give my place of birth, or home town, etc. etc, that's also part of me, and just as irrelevant to whether I'm old enough to buy alcohol.
90490
Kevin Unhammer
Posted about 3 years ago
David Birch: A new way to stop identity theft
But Verizon wouldn't hold the database. The only way they could get more info on you than you gave them was if they installed spyware into your phone without your knowledge. Actually, AT&T, Sprint and T-Mobile did install spyware (but strangely not Verizon), a program called CarrierIQ which, among other things, recorded and uploaded keystrokes (such as "passwords to otherwise secure websites"). But I guess if your private key was only used in a Trusted Execution Environment, CarrierIQ shouldn't be able to get a hold of it, nor the output of authorisations, as long as there aren't any bugs … I guess exploits and spyware are always a risk. There would have to be heavy sanctions against any kind of CarrierIQ-like behavour with PychicID.
90490
Kevin Unhammer
Posted over 3 years ago
Yochai Benkler: The new open-source economics
In the bitcoin system, the coins are all "visible". To use the bitcoin analogy, say a coin was a book. I have a coin/book, and I transfer it to you. What in fact happens when I transfer it is that I send out a message to the network saying "node X is transferring this exact coin to node Y". The network spreads out this message, and when enough nodes have confirmed the message, the transaction has happened. But I can still "read" the coin, even though you own it. I just can't use it, because everyone in the network have agreed that now you own it (if I try to send out a similar message again, it won't work).
90490
Kevin Unhammer
Posted over 4 years ago
James Geary: Metaphorically speaking
When they study metaphors in cognitive psychology / cognitive linguistics, it's the "commonplace" metaphors that are of interest. If you use a metaphor "as a metaphor" (or, precisely because it's a metaphor, because you want a literary effect), that's less interesting to the linguist -- what's interesting is how deeply metaphors permeate _regular_ speech. E.g. how I used "deeply" to describe a relation to the completely abstract concept "speech", which of course can have no physical depth, not being physical. Also, to what extent (if at all) does this deep integration of metaphor into regular speech shape the way we think.
90490
Kevin Unhammer
Posted over 4 years ago
James Geary: Metaphorically speaking
In Latin, a g is pronounced as in "got", not as in "jot". I liked the talk all the way up to the "shake things up" part. That just doesn't make linguistic sense. If Descartes really meant "shake things up" when he wrote "cogito ergo sum", he wouldn't have been going on about Thoughts etc. in the rest of the text. When people use words, they typically don't think about their etymological history (if they even know about it). If we did that when we spoke, we'd never get anywhere. (Oh no! I used the morpheme "-where" which etymologically has the same proto-indo-european root as "who" so I must've meant "we'd never get anywho" which I'm sure you don't have to be too twisted to construe sexually. Whatever will I do...)
90490
Kevin Unhammer
Posted about 5 years ago
Jay Walker: The world's English mania
This kind of "One Language" idea is scary. Remember the "Welsh Not" custom a hundred years ago in Great Britain, where students speaking Welsh would be given a lashing by the teacher. Similar punishment and humiliation happened to Sámi speakers in Norway, Gikuyu speakers in Kenya (where English was enforced), Ainu in Japan, and so on and so forth.
90490
Kevin Unhammer
Posted about 5 years ago
Jay Walker: The world's English mania
"It is going to survive and it will become /the/ Universal language because of the Internet and mass media. " Actually…no. Machine Translation will become the universal language. Of course, it'll distort meaning and sound funny, but then so does globish.
90490
Kevin Unhammer
Posted about 5 years ago
Jay Walker: The world's English mania
IN NORWAY CHILDREN ARE REQUIRED BY LAW TO LEARN NOT ONLY ENGLISH BUT MATH! IN THE FIRST GRADE! He really does make this government-enforced Education sound awful. I have to say, the fact that this talk got into TED is quite disappointing. As has been commented below, the clips are from a special school in China called Crazy English (ie. not everyone takes the screaming class) and China ranks rather low on the list of population with English as a second language. Also, I really don't see what his argument is. Does he really believe English is somehow better-suited than other languages to be the lingua franca? (Latin had that postion at one point, French did too; not for linguistic reasons, but political ones.)