591 karma · joined June 2, 2016
That's an enormous "if". Fusion research has been underfunded for decades, we don't even know if we can sustain it, and for a reason I don't get, many ecologists are anti-nuclear. Plus fusion won't solve everything. You won't put a sun core into your car nor your plane, not in the next centuries.
> Even if we don't, the plummetting cost of wind and solar may solve that anyway (by "solve" I mean that as soon as non-fossil fuel power production is cheaper than fossil fuels it becomes economic to remove CO2 from the atmosphere and turn into hydrocarbons).
Oh, you mean the unreliable power sources that sometimes provide electricity, sometimes don't?
> The Earth has also been a lot warmer than it is now
Well yes but 5°C is also the difference between an ice age and a hot age. We're currently in a hot age and the 4°C increase we're heading to are not very engaging. (Remainder, during ice age North America down to NY and Northern Europe down to Paris were covered by enormous glaciers. So what will the climate be in 4°C?). https://xkcd.com/1379/
Also in human history the Earth has never been this hot.
> Too expensive to get to space? Eventually the cost of this will go down to dollars per kilogram.
Ha. To escape Earth gravity you still need to spend 6.2*10^7 J/kg. That's gonna be a long time before we get to 1$/kg given that rocket propulsion is extremely costly (and as far as we know can't rely on atomic energy either). We're still at ~25 000 $/kg. And anyway recent studies show that the human body is unlikely to bear a trip to Mars, let alone the health issues caused by weightless environment.
Edit, source: http://www.pnas.org/content/early/2018/09/26/1807522115. Basically the study suggests heavy ion irradiation is dangerous for the digestive system. The only thing is I'm not sure about the number of mice they used.
> We also seem to have a pretty poor history of predictions when it comes to climate change too.
I mean it's not as if oil companies conducted their own research and reached the very same conclusion in 1977. Oh wait. https://www.scientificamerican.com/article/exxon-knew-about-...
I'm sorry but you're just burying your head in the sand, making the problem someone else's.
Also there's no way the rest of EU is going to accept for UK to stay in the EU with the same privileges as previously (non use of euro, tax exemptions...). Especially since UK was one of the most critical members, now EU has more incentive to move forward.
Personally I can't see how UK can rejoin the EU without applying from scratch like Serbia and so on.
So either the EU decides to submit to the US or decides to grow a spine. It appears it's the latter. Too bad for the US.
It's good to see that Europe finally grew a spine.
[0] http://www.lefigaro.fr/flash-actu/2014/12/04/97001-20141204F... [1] https://www.lesechos.fr/06/12/2015/lesechos.fr/021536423616_...
Copyright inheritance is not taxed in any way.
He takes a very strict stance on copyright law, and does not give a shit about parody, art or even hommages. Moulinsart will take every opportunity to strike down your Tintin webpage, even DMCA takedowns even though you're French and Moulinsart is Belgian (see what happened to [0], an explanation is available in French in [1]). Rodwell also wanted to publish a new Tintin episode, just to prevent the license to fall into public domain.
The guy has all powers on Tintin, because he married the author's widow. And that's it.
[0]: https://lepetitvingtetunieme.tumblr.com/
[1]: https://www.lesinrocks.com/2014/03/26/actualite/tintin-la-ma...
I'm a bit doubtful about this point. My experience (2 4-months internships inside Paris, lived in Paris another year but working in suburbs) is that restaurants are more expensive than the ~8€ Ticket Restaurant. Never ate for less than 12€ in a bistrot, and obviously for this price you only get main dish, no desert. In one company we had this hot food delivery inside our walls that would cost exactly one Ticket Restaurant, but nothing similar with bistrots.
To be honest one of my internships was next to Bir-Hakeim (metro station closest to Eiffel Tower) so the prices may have been inflated, but still.
Biggio et al., "Evasion Attacks Against Machine Learning at Test Time", http://www.ecmlpkdd2013.org/wp-content/uploads/2013/07/527.p...
However I'm pretty sure this was advertised or at least acknowledged by Google in the launching of reCAPTCHA v2.
So while your objection is valid, in practice any truly randomly selected real number is non-computable.
Plus, life took about 1 billion years to emerge. This is huge. The solar system took about the same time to form. Also, life may be extremely slow to appear, but once it's there, progression is quicker. Since the beginning of phanerozoic eon (and Cambrian explosion), a geologic period (Pleistocene, Jurassic, Cambrian...) only lasts a few tens of millions years. I find it very plausible that at the beginning of Cambrian era, a critical step in evolution was reached, allowing a better and faster evolution.
Secondly and most importantly, GDPR is not about preventing people to get information, but rather about preventing people to track customer. If I use your example, GDPR would prevent you from following someone after they watched you juggling. It would prevent you from following them home, noting where they work, who are their acquaintances, what kind of food they eat and so on: it's technically legal but incredibly creepy. If someone were doing this to you you would be the one to tell them to piss off. You can perfectly write what are the observations of the passer-bys who watched you, as long as the log is anonymized. Which is easy to do and not harming for your business. You didn't want to track them anyway, did you?
(a) is a classical side-channel attack, and as I say non-homomorphic libraries already take these kinds of attacks in consideration. HE won't be an exception to that, depending on what level of privacy is needed.
I did not understand what you mean by (d), and to be honest by (b) either. If you have any papers/blogs about that I'll be glad to read them.
And yes I agree, successful FHE system is inexistent as for now. But the evidence of the bare existence of FHE is only ten years old, somewhat practical FHE is even younger. We've not even reached practical FHE yet, security issues will be tackled when they will become the blocking point. At this point of the technology you can't expect research to be immediately applicable. ZK proofs were first designed in the 80s, and as far as I know they only started to be used in practice (zk-SNARKS for instance) recently.
About your query example, (F)HE never claimed to be secure against leaked information inherent to the query. This is more the resort of privacy, which is tackled by differential privacy researchers (and as such has nothing to do with FHE). Naturally you can put differential privacy on top of FHE but we're not there yet. The biggest problem here is that crypto is not magic: in your example, NO system can truthfully answer to the question without leaking the only male's grade, crypto or not. So maybe you'll restrict the system to run the query only when there are at least two male students. But then you're still leaking info to me if I'm the second student, and so on. Wanting crypto being able to do this kind of job is equivalent to wanting cars able to go fast anywhere anytime but that should never crash into a tree at more than 5 km/h: if you can go fast anywhere, you can go fast into a tree. Sad but true.
About the ability to subtract and compare, I don't understand what you are talking about, could you explain it?
About ECB, I mean, yeah, everyone in crypto knows ECB is insecure, this is stream cypher 101 and equivalent to "ROT13 is not secure, do not use it for crypto". But this has nothing to do with FHE.
Keeping track of time is a side-channel attack. Nothing to do with FHE in particular, it's common to all crypto. Most crypto libraries are secure against timing attacks, as it is one of the most obvious side-channel attack. FHE libraries will hopefully be too, just give them time. Also I think FHE is by nature more resilient to timing attacks as it must natively implement simultaneous computation of if/else branches, for instance.
Long story short, yeah, crypto is insanely hard to implement correctly. When your problem is well-defined (first difficulty), you must find the good tools (second difficulty), with the correct security parameters (3rd), and implement them correctly: no bugs (4th) and no side-channel attacks (5th). That's the reason why the first rule of crypto is "don't write your own crypto".
I feel that the article is a bit vague about the applications of homomorphic encryption, so I'll add some examples. For instance, suppose that your mail server encrypts your mails, and only you can decrypt them. If you want to search something trough your mails, you have two choices:
* download all your emails, decrypt all of them, run your search, discard the rest
* allow the mail server to decrypt your mail, and let it run the search
Both options are not viable. The first one is extremely costly in bandwidth, storage and computation, the second one contradicts the premise of encrypting your data in the first place. (Fully) homomorphic encryption would allow the mail server to perform the search on encrypted data, not knowing the content of the mail, or even the query for that matter.
Another example, you want to get a loan from the bank. Their conditions are: not having served jail time, no critical health state, and no gambling addiction. Further more they allow you to check one of these cases if you are a veteran (go figure). All these information are sensible private data, which the bank could use to rise your interest rates. Yet, you need the loan. Using homomorphic encryption and some variants could allow the bank to only get the result: whether you are eligible to the loan or not.
Of course, the bank would have to agree to such a scheme in the first place and what I describe is still highly theoretical but I hope you get the idea of FHE's potential uses.
------
I'll also add a few precisions about the article, several homomorphic schemes exist, the one described in the article is called Ring Learning With Errors (RLWE), a derivation of the LWE problem. Mathematically it boils down to solving a problem called shortest vector problem in a lattice, it's a bit abstract but you can Google it (basically, you have a regular grid of points, find the smallest vectors that can generate all points). It's an NP-hard problem, but you have to pick your lattice very carefully (many instances are solvable in polynomial-time).
Also, if I'm not mistaken, the encryption system described in the article is a Somewhat Homomorphic Encryption (SHE), which means that you cannot use it for complicated operations. Well, the paper it's based on is Fully Homomorphic Encryption, but the techniques to reach it are not given in the article. After a while, the errors become preponderant and you cannot decrypt anything any more. Several schemes actually solved this issue, reaching Fully Homomorphic Encryption (FHE). The first FHE scheme was described by Gendry in his PhD thesis in 2009 (damn I wish my PhD will be half as good as the quarter of his). Yet if I remember correctly many FHE schemes are still unpractical (I remember a paper in which the polynomials were several Gbytes). But the field is quite young, and research is progressing!
Edit: however now that I look into it, I should rather have used "conditional probabilities" name rather than Bayes'. I always mix both of them, my bad.
Your neighbours has two children, one of which is a girl.
What is the probability of the other child to be a girl?
A. One may say 1/2, but actually, using basic Bayes theorem: P(2 girls | at least 1 girl) = P(2 girls AND 1 at least girl) / P(at least 1 girl) = (1/4) / (3/4) = 1/3.I think this problem is actually equivalent to the Monty Hall problem. But it can get much weirder.
Your neighbour has 2 children, one is a girl.
You know the other child is born on a Sunday,
what is the probability that this other child is a girl?
A. Similarly, counting all possible cases, one does not get 1/2 or 1/3, but rather 13/27.