cba9

Although for a simple task like this demo, I stand by simple conventional neuroevolution as they would work best and are easy to train as the nets are quite small.

They may be easy to train but I'm not sure about 'work best'. Watching your final trained agents, I see a lot of dumb obvious mistakes where the agent runs right into a line; and I see little higher-level strategic planning like trying to maximize free space around one or run away from a incoming line.

To some extent, you probably have heard of them, under the rubric 'direct transfers' or cash payments. Direct transfers of valuable assets (cows or cash) to the very poor have done very well when evaluated in randomized experiments, and I suspect you've seen at least one article on them in the past (albeit it might have been on GiveDirectly or one of the other programs rather than BRAC itself).

Evolutionary algorithms may work ok here, but I have not seen any high-performance ML tasks done with evolutionary training of neural networks in a long time. It's all backprop and, increasingly, reinforcement learning these days.

Have you ever looked at https://github.com/karpathy/reinforcejs ? Karpathy already has demos similar to yours which you could build on. The trained agents might be better.

Historians believe Suleiman’s heart and internal organs were buried in the tomb and his body taken back to Constantinople, as Istanbul was then known. His death at Szigetvar was kept secret for 48 days to prevent his troops from giving up the fight.

I was wondering how the tomb of the most famous sultan could have possibly gotten itself lost. The answer is it's only sort of his tomb.

That's too elaborate. They like more direct attacks. Think subpoenas exploiting the third party doctrine to the hilt, MLATs to image any server remotely suspicious outside the USA, and occasional direct hacks.

Source for all that?

So it's difficult to successfully get it, and there's a lot that could go wrong and, say, prompt a lawsuit or criminal case?

http://www.businessinsider.com.au/the-australian-who-may-hav...

"Under the scheme, a company which turns over less than $20 million is eligible to apply for a cash rebate of 45 cents for every dollar spent on research and development activity.

According to a press release in May, 2015, DeMorgan Ltd succeeded in its application for the rebate.

That means DeMorgan were able to prove to AusIndustry that it had spent $120 million on R&D activities in the 2014/2015 financial year, all while turning over less than $20 million. A 45 cent rebate on $120 million works out to $54 million."

wouldn't Wright have a huge income tax debt?

Yes, but presumably he can pay it out of his Bitcoin fortune. He has to deal with that at some point anyway, and by going public now, he gains additional leverage on his VAT or tax credit or whatever the ATO has decided to get him for this time.

At this point, if he is Satoshi, he would probably be better off coming out of the closet. If you read through the McGrath-Nichols documents about the Hotwire bankruptcy and how the R&D tax credits dispute killed Hotwire, it's clear the ATO has it out for Wright.

I bet that whatever this raid is about, it's about something like R&D tax credits again (http://www.businessinsider.com.au/the-australian-who-may-hav... says $54 million in 2015 alone!) or some sort of VAT rebate, and it may kill another of Wright's companies. There's not much you can do against the ATO if they are determined to reach an adverse decision against you... But if he's Satoshi, he can take the moral high ground and wage a media campaign to shame the ATO into dropping the case and maybe finally correctly treating Bitcoin as money.

So let’s recap: in September 2015, Wright was annoyed that people didn’t know who the real Nakamoto was.

Is Kashmir unable to read? Wright never said that. He was clearly angry about the national stereotyping and the 'assumption' and how thinking is 'limited'. This is exactly like those cases where women authors publish books under names like 'J.K. Rowling' and readers assume they are men; of course they are going to be upset.

And Wright lives in London according to that Vegas video. Even more billionaires live there without much worries about kidnapping, and you'd think the Russians there would be a testcase.

Apparently Gizmodo has been investigating this for even longer (!) and has a lot of other evidence for Wright=Satoshi too: http://gizmodo.com/this-australian-says-he-and-his-dead-frie...

So your point is that capitalist countries became much wealthier and they also did this much earlier than protectionist countries did, and this is somehow a bad thing about free trade?

I don't see any evidence that the cycle has gotten faster. There were plenty of mayfly wonders on YouTube years and years ago, and the Internet had its memes well before. It doesn't look much different to me over the past decade, (Have aggregators like Fark, YTMD, SomethingAwful, 4chan, really sped up? The list is so long: https://en.wikipedia.org/wiki/List_of_Internet_phenomena If we looked at knowyourmeme.com's lists and the Google search histories, I don't think we would see much of a speedup. Memes would still be born and die on similar timescales.)

...all of the countries which have done extremely well (China, Brazil) have protectionist economies while the countries that have stagnated or suffered (US, EU) have fervently implemented free trade policies....Empirically, countries that have not put all of their eggs in the proverbial free trade basket have been much more prosperous than those that have.

Are you saying that Brazil and China are wealthier than the US and EU?

I also think that this is a curious way to describe China's recent history in going from Mao to Deng...

I wondered that too but then they cover how it's been embroiled in a legal battle for 30 years now. So apparently its location more or less was discovered a long time ago, it's only now that the legal battle has ended that more detailed investigation can take place.

Is it just me, or is the "ANALGESICS AS RISK FACTORS FOR F-AD: (2) EPIDEMIOLOGY" section really uncompelling and the epidemiological evidence seems to amount to a few inconsistent cross-sectional studies with no particularly large correlations like implied, and a note that both paracetamol and Alzheimer's have increased over the past two centuries or so?

The article is not well written, and I personally had to parse it several times to figure out what he was trying to say. I'm still not even sure if this is the correct interpretation.

I thought it was perfectly clear. He was telling a funny story about how systems and technologies evolve, giving two examples of that (latter, the watch, former, the system's airgap springing a leak), and furnishing an object lesson in the need for regular thorough audits to ensure that systems and controls thereof are still in place and still working the way that the owners think it's working.

A facility like the one he described should have had regular security audits to verify that no hard lines were placed where they should not be.

Exactly. In fact, I believe at the time he wrote this blog post, OP was an active auditor for BDO. In some of his other posts, he analyzes observations he made while auditing a variety of companies/organizations; unsurprisingly standards across the board are very poor. He would be the first to say that this sort of thing is what an audit should prevent and why audits are needed (although I'm not sure I agree with his venom against pentesting; which I see analogous to fuzzing).

The attacker uses the cracked 2005 stamping key to stamp the file and the cracked certificate to sign it.

OK, so you don't actually understand how cryptographic timestamping works or the Bitcoin version. Please read up on it before commenting.

If you could certify a timestamp forever, in a non-expiring, uncrackable way, why wouldn't you just use that for the whole file?

What you are timestamping is the whole file!

Perhaps you should read the submission. The larger point here is that even if you set up a network in the first place which is genuinely airgapped, as time passes and systems evolve there will be constant pressure from within and without to re-establish a network connection somewhere in order to make everyone's lives easier and eventually, whether deliberate or inadvertent, a connection will be made (and of course, we know that the NSA has a variety of infiltration and exfiltration methods to get across air gaps, such as dropping flash drives and waiting for an insider to be foolish enough to bring it inside).

Believing that an air gap exists or will continue to exist indefinitely is hence setting yourself up for some unpleasant surprises in the future, and encourages weak security designs where the network/system is crunchy on the outside and all delicious and soft and gooey on the inside. (Which is more secure, to have your local WiFi set up with WPA or whatever and have employees telnet into servers, or just go Google-style and have fully encrypted end to end links without requiring any belief in security of the links?)

Good question - I have no idea. But if I google 'gas-filled tube ethernet', I get a number of hits like http://en.tdk.eu/blob/174150/download/5/smd-surge-arresters-... and http://www.first-electronic.com/uploadfile/2010989552637708.... which suggest that these products are sold for networking purposes to prevent surges. Since this is in a military context, I would hazard a guess that this may be some sort of standard hardening requirement to try to protect the datacenter against lightning strikes, EMPs (such as from nuclear strikes), and general accidents. This may sound paranoid, but then again, so do Faraday cages, and it is the military - it's their job.

I did not know that about Japan, but apparently here, as with the fax machine, Japan is an exception. Japan may spend $500m on telegrams (which incidentally, doesn't seem like very many telegrams per person per year - if each is cheaper than I was quoted by quite a bit, say $50 total, and there's 127m people, then that's one a month on average. I'm also not sure they spend that much since I can't find any quick hits in Google discussing it recently; perhaps you could blog about it sometime.), but nevertheless what delivery services in Japan do or people do in Japan is not that useful, in terms of amortizing fixed costs and realizing economies of scale, to American or Australian delivery costs, nor is it informative about how often Americans or Australians send telegrams.

I also didn't bother to check Fedex/DHL because I figured while they probably did have faster deliveries than USPS, they always charge way more and we were unsure we really needed within a day delivery (which is the only place on the Pareto frontier the private companies would be for general document delivery).

As it happened, we were unsatisfied with our options so we kept digging and eventually came up with what we think are the phone number & email address of the person we need to reach, so hopefully it all turned out to be moot.

If the key is cracked, then the attackers can create a brand new file, and make it look like its existence had been certified in 2005.

What? If someone cryptographically timestamped a particular signed file in 2005, even if Eve cracks the original public key and can now produce arbitrary new signed files with that key, they still do not have 2005 timestamps for any of those new files! They could only make 2015 timestamps. That's the whole point of cryptographic timestamps.

Right. At one point in the complaint, they say the company had oversubscribed its mining capabilities but kept selling to 4x IIRC, which rather implies that there was mining capability there in the first place (just much much less than they were selling).

The problem is you don't need fast physical deliveries of short messages that often. For example, a few days ago I was pondering how to reach a particular person on very short notice; I had an address, but nothing else, so I couldn't call them, and while I could spend like $50-100 on international mail, it wouldn't get there faster than 3 days from sending. Ultimately, the best solution I could find was... telegrams. Specifically, their modern email+courier incarnation. It would cost like $70 and probably get to the person in a day or two.

I thought this sounded outrageous, but then it occurred to me: when was the last time anyone ever used telegrams? There can't be many uses of them - I've never used one in my entire life - and all the regular costs of a business still need to be paid and the couriers have to be paid and so on.

How pleasantly surprising to see an official link which is the fulltext as well. I'll just quote the abstract to entice people into reading it:

"The anthropological record indicates that approximately 85 per cent of human societies have permitted men to have more than one wife (polygynous marriage), and both empirical and evolutionary considerations suggest that large absolute differences in wealth should favour more polygynous marriages. Yet, monogamous marriage has spread across Europe, and more recently across the globe, even as absolute wealth differences have expanded. Here, we develop and explore the hypothesis that the norms and institutions that compose the modern package of monogamous marriage have been favoured by cultural evolution because of their group-beneficial effects—promoting success in inter-group competition. In suppressing intrasexual competition and reducing the size of the pool of unmarried men, normative monogamy reduces crime rates, including rape, murder, assault, robbery and fraud, as well as decreasing personal abuses. By assuaging the competition for younger brides, normative monogamy decreases (i) the spousal age gap, (ii) fertility, and (iii) gender inequality. By shifting male efforts from seeking wives to paternal investment, normative monogamy increases savings, child investment and economic productivity. By increasing the relatedness within households, normative monogamy reduces intra-household conflict, leading to lower rates of child neglect, abuse, accidental death and homicide. These predictions are tested using converging lines of evidence from across the human sciences."

As so often with Dali - /not sure if serious or trolling

Twitter: https://twitter.com/cmdrtaco/status/670424137282777088

Generative models strike again.