enki

not if the amount of training data required to learn by example is exponential or even combinatorial. we might not even be in the same ballpark. all data ever recorded on digital media might not be enough to learn even a simple intelligence without assumptions about structure.

a kind of analogous problem is: can you learn how to build a living thing purely from digitally recorded DNA samples, if you don't have access to the internal structure of any living thing? How many dna samples would you need?

as per http://paulbohm.com/pdf/10.1.1.76.5543.pdf and http://paulbohm.com/pdf/general_limitations.pdf learning by example requires many more examples than the kolmogorov complexity of the target concept.

T then is not just the life experience of one human - to learn from observation by example, to reconstruct a human, you'd need to observe the life experiences of trillions and trillions of humans to gain sufficient information about humans to narrow down the possible implementations that match human behavior not just mechanically given the same input, but also given new previously untested input.

at 3gb encoded as dna the search space already is huge, but that ignores that the genome alone doesn't contain the information needed to read it. (e.g. you need a living thing to use the DNA, for it to make sense)

author: there's no fundamental objection. i'm only pointing out that throwing faster computation at the problem doesn't necessarily solve it. right now all the work essentially needs to be done by humans.

If you follow the link to the lower bound to learning by example, you'll see that it's really hard to learn by example if you haven't pre-encoded ideas about the interpretation into the reading program. Without prior knowledge, you need lots and lots and lots of examples.

All the information that humanity has ever stored on digital media is likely not remotely enough to reconstruct a human mind from.

That help won't be coming from politicians, because they don't even remotely get that wealth and wealth creation are linked. And despite what OP wrote, the fundamentals haven't changed in any other way that was perceptible to me.

I can't tell you exactly what's wrong with startups in Europe. It's no one thing. It's thousands of things.

Unless you've experienced the Bay Area, it's probably impossible to anticipate what you're missing. I didn't. The stuff you get in the Bay Area is stuff that doesn't transfer over the internet - so just by reading hackernews and techcrunch you have no means to figure out what is or could be different.

I think Paul Graham's argument (paraphrased) that all startups die by default, but in some environments they sometimes get saved, might come closest in describing what's happening.

I think the statistics argument really is the most solid one (though admittedly self-fulfilling, because it means the most driven entrepreneurs leave). Unless something fundamentally changes, any improvements to the ecosystem will have to be gradual over a long time.

I'm far less optimistic. I was one of the most vocal and active pro-european-startup activists before I left for San Francisco in 2009. With no capital, I started multiple companies. Some failed, some are still active and now do combined multiple millions of annual revenue (e.g. mjam.net). I started Europe's first YC-inspired accelerator (YEurope) in 2006 and helped put the current global hackerspaces movement into motion so we'd have support infrastructure. Back then I regularly talked to european politicians up to the president level.

The point where I left for San Francisco was when I realized that no one there realizes that there is a causal link between wealth creation and having wealth. They don't know they have a problem, or even should they suspect it, they don't know what causes it.

On a recent trip back I haven't noticed any change in that - so even if they wanted to improve something, they wouldn't know what or how. The economic crisis provides a convenient excuse for all problems, not an incentive to explore what's wrong.

The easiest argument against Europe for startups is sadly on an individual level: Statistics. If you're one of the best - but not the best - entrepreneur in Europe (500 Million People), you know that almost no one more talented than you ever succeeded. And almost no one less talented either. In San Francisco, both people more and less talented succeed all the time.

So statistics. It's far more pretentious to think you're so much better than the best among 500 million, than to think you'll do well among the 8 million in the Bay Area.

i love your description of this.

using up resources is not valuable by itself. forcing someone to spend on anything, anything at all, does not lead to a good allocation of resources. it might increase GDP, but it does not create value.

under inflationary interpretations of value creation, building a road circling back onto itself - not connected to anything - in the middle of a desert that no one ever visits - increases GDP and thus helps the economy. the better if you afterwards tear it up and rebuild it a second time.

on the other hand, in a stable currency system, if someone holds on to their capital, they actually free up resources for those who do spend. this means the purchasing power of those who do spend goes up.

in such a system money flows away from those who spend on things that have no return (consumption), and towards those who create (value creation by seizing and creating opportunities).

only if you're ok with never buying any physical goods with bitcoin. the weak point is the exchange between bitcoins and classical currencies or physical goods.

thanks! date added to template!

How is bitcoin a non-commodity currency? If bitcoins have value because they allow you to do things you couldn't otherwise do (my point), it's a commodity.

why is it desirable for a currency to be a non-commodity currency?

bitcoin's problem is that the financial sector is massively regulated - you're going to jail if you attempt to compete.

creating a website where you can store $10, send it to me, and then let me withdraw it, is a 5 year felony in california. that's even when the site is complying with all anti-money-laundering and know-your-customer regulation that could be argued for because of terrorism. it's purely a licensing issue.

bitcoin's influence probably won't significantly grow until people find legal workarounds/ways to comply (of which there are plenty, but likely implementing them takes someone committed at least a year), or something happens that makes enforcement of the law unlikely.

i personally know of a sizable number of bitcoin startups with working/finished software and a great marketing strategy, that gave up in the face of financial regulation and threatened jail time.

the dollar and euro are also assets in that way. you're usually just not aware of its fluctuations (against each other, against other currencies, and against commodities)

look at any of the US dollar indexes, which examine the exchange rate of the dollar against other currencies to shake the belief that the dollar is stable: https://www.google.com/finance?chdnp=1&chdd=1&chds=1...

because an attacker controlling less than 50% of the computational resources can be lucky and find the first solution before anyone else.

he can even be lucky twice or thrice. but to be lucky 12 times against everyone else with less than 50% of the resources, is statistically improbable.

2 hours is the result of those two values (12 being statistically relevant, and 10 minutes a chosen parameter of problem difficulty), not an input variable.

yes, bitcoins in theory can be infinitely divided. classical observations about hoarding and "bad currency driving out the good" likely don't apply. if you want to spend, you just spend however little you want to spend.

author here: there's a cost with storing gold in the ground and securing it, and there's a cost with printing counterfeit-resistant paper and finding and jailing counterfeiters (and a cost with not doing so - unbounded inflation)

an interesting question then is: are the above costs greater than the cost of operating bitcoin? it might seem so at first, but don't be so sure and run the numbers on the back of an envelope.

also walk around downtown in SF and NYC and look at the amount of skyscrapers that are nothing but banks, scratch your head, and ask yourself what value they're actually adding.

author here: i've written 12x10 minutes, and the original description writes 12x10 minutes. is there something i've missed?

it takes 10 minutes to find a solution, but only after 12 solutions have been found (the chain has been extended 12 times total), can the authenticity of the first of those 12 be guaranteed.

analysis: http://golem.ph.utexas.edu/category/2012/07/zerodeterminant_...

this should help: http://en.wikipedia.org/wiki/Interactive_proof_system

A hip kid investing in things he likes probably has a higher hit-miss ratio than most VCs... :)

i retrained myself to touch-type when i was 16.

only thing it took was forcing myself to not type any other way, no matter how slow i ended up being. within a month or two i was typing at my previous speed. (and continued to improve from there. these days i type faster than almost anyone who isn't training for competitions)

no tutorials, guides, books, teachers needed. just do it.

ps: yes, those two months are painful. try getting into a heated argument on irc and losing because you can't type fast enough. don't break your touch-typing-only rule. just try hard to type faster the right way! that's where the motivation comes from ;)

even if you had had the same brilliant insights into the graph structure of the web when they did, you most likely would have failed because it was prohibitively expensive (the cost in the article is probably underestimated by orders of magnitude). it's simply a fact that:

1) getting the data, 2) computing the eigenvector of a large matrix, 3) and serving that data to users, wasn't cheap in 1998. it's comparatively dirt cheap today.

not to diss larry and sergey's impressive achievement - they were brilliant and they pulled it off - but i think back then game was so costly that a lot of brilliant people never made it to the starting line. it's cool to see that it's become a much more level playing field now. i'm curious what cool stuff we missed out on because of people who didn't make it to the starting line!

"The SDLP is notable in the history of Google as a primary sources of funding for Lawrence Page's and Sergey Brin (Brin was also supported by a NSF Graduate Research Fellowship) during the period they developed the precursors and initial versions of the Google search engine prior to the incorporation of Google as a private entity"

This included a $4,516,573 NSF grant (that didn't go to Larry & Sergey in full, but probably helped their project's infrastructure quite a bit).

http://www.nsf.gov/awardsearch/showAward.do?AwardNumber=9411... http://en.wikipedia.org/wiki/Stanford_Digital_Library_Projec...

On the expense side I've probably actually underestimated the expenses by orders of magnitude. Bandwidth wasn't cheap back then and the storage requirements probably were significantly higher.

In http://www.salon.com/1998/12/21/straight_44/ it said "Page says the current version of Google, which has indexed about 60 million pages, will continue to be improved as the company expands." and http://en.wikipedia.org/wiki/History_of_Google#cite_note-sal... said Total indexable HTML urls: 75.2306 Million Total content downloaded: 207.022 gigabytes.

You missed the temporal context of what I wrote:

"A few decades ago [...] if you'd wanted to use a hash table, if you even knew what a hash table was, you'd have to write your own."

I agree with you on the quality of the BSD code, and I'm glad such great code is readily available. But a) I definitely had been programming before 1990 (the copyright on that BSD code), b) back then hash tables were far less tightly integrated with programming languages than they are today and fewer people knew about them, and c) if you want to be pedantic Hash Tables have been around since 1953, so way before most programming languages that are still in use today. http://en.wikipedia.org/wiki/Hash_table#History - they're however much more commonly understood, and in ubiquitous use today!

good point! google definitely is part of what makes us so much more productive programmers today! (despite google codesearch going away)

sorry, we didn't get to making a better site yet, but we'll definitely address your concerns when we get to it!

(author here)

I was writing this more in the sense that kids at BioCurious (and the DIY Bio Movement in general) are doing electrophoresis to transfer DNA from glowing jellyfish to bacteria. This is just a few (two?) years after someone got a Nobel prize for that.

That's progress. If stuff that used to be hard falls into kids hands, you're gonna see impressive stuff happening.

However I fully agree that it takes more than that to build a company (Also I wouldn't try to compete with 2012 Google using 1998 technology)