izendejas

edited: to respond to abraham's post directly

There's a difference between training with that data and testing against it, which is precisely what I wanted to highlight here since many expected Google to use such data for training. You can't do that since you can overbias/over-fit given what little daat we, plug-in users, are able to produce versus the amounts they have access to.

2nd edit: it should also be noted that that 84% figure they mentioned related to the extension data is a measure of recall. Assuming Google wants to do a better job with precision in these cases (don't want to get many false positives), that recall is still fairly good in mind since I'm sure there's some noise in that data.

Brilliant use of crowd-sourcing to gather a test set via the chrome plug-in. Makes absolute sense, otherwise you can overfit. Great turn-around time also.

As someone who is looking to apply, even at this stage, having YCombinator as an option is tremendously invaluable. So on behalf of the dreamers out there, I say thanks Paul, Jessica and crew. Even if not accepted, just having that door there gives one more clarity moving forward by removing lots of the uncertainty.

You need to check out Freebase.

I think this is only a problem if you're targeting niches because as I see it, if you can't use your own product while pretending to be your non-technical mother, then you shouldn't be in business unless you're selling for a premium. Unfortunately, technical geeks won't pay a premium and they themselves can often build something relatively useful with, in this case, even a simple google search: http://www.google.com/search?sourceid=chrome&ie=UTF-8...

I actually went through this exercise before and found it to be annoying, but how often do you search for and buy books? It's not everyday you do.

Don't get me wrong, I've bookmarked your site and find it useful, but I probably won't go back to it in a while.

I agree. The article I link below (and which I believe I first saw here on HN) should be closer to the model of what I would hope to see more of around here.

It gives you easy to read data which I believe is often more helpful than any opinions because at the end of the day you want to know, what are the trade-offs between performance and development time*, among other things. And it's also careful to point out many relevant caveats.

http://faves.eapen.in/performance-comparison-c-java-python-r...

Right. That's what Schmidt is referring to when he says Google wants to do "search for you".

Assuming Google makes privacy controls more visible and there's someone to keep them in check, I frankly believe that this is a better model than what we currently have today. Many of the ad companies are way more evil (think spyware) and way more ineffective. In their mind, they can't be transparent. But that's because they want to take shortcuts and track your every click surreptitiously without really wanting to add any value to you directly (that is they don't have as big a stake in getting it right), the user, and only to turn around and sell this data to others and make a quick profit. That model is broken from the beginning and I'd rather not continue down this path.

So if Google, Facebook or anyone can give me more controls, more transparency, then that to me is a more compelling model and I'll gladly share more relevant information with them because I know that they'll turn around and offer me better search results, better ads, better book recommendations... whatever. If they don't someone else will, but at least with Chrome OS, you have the option--don't buy a machine with it.

It seems to be machine-dependent? I remember opting out a while back at work, but my mobile phone definitely shows me a profile, albeit one that isn't totally accurate.

In any case, you guys should also checkout Bynamite. They've got this interesting mission to give users more control over what advertisers see. I'm all for that not just from a privacy standpoint, but also from a relevancy one.

Don't miss the link at the bottom. Yahoo! and Facebook will present the next gen Hadoop this coming Wednesday (Feb 16): http://www.meetup.com/hadoop/events/16116942/

-- If you're attending and your interested in data mining and personalization, please look for me. I'd love to chat. twitter: @izendejas linkedin: http://www.linkedin.com/in/izendejas

"AI is a software problem and if you don't know how general intelligence works, having a new kind of transistor really, really, really ain't gonna help you much."

I'm not sure anyone is arguing against the latter part of your statement but you as you contradict yourself by arguing that it's a software problem first. AI is a biological, psychological, cog sci, EE/hardware, et al problem because, as you state, understanding how general intelligence works is important--as it is important to know how to build the technology that will run it (ie, possibly the memristor + something else, of course).

Sure you can program some sort of AI, but try doing it at the massively parallel scale of a brain. What I think the memristor opens up is the very ability to do this. I'm not exactly sure how--not sure that anyone does yet--but "in-memory" processing and moving away from the von Neuman (serial) architecture might be a good step as I'm not sure we can easily develop highly parallel software without letting the low-level architectures handle that automatically in more of a stochastic way.

I'm staying healthy as long as possible. An exciting future awaits.

Forget going to Mars... the next frontier is understanding and emulating a mammalian brain, preferably the human brain with its amazing cerebral cortex. And boy will the memristor and related technology come at a great time when we're bombarded with an exponential increase in data that we don't yet fully exploit.

This has potential. I asked for this very thing about two months ago and surely someone is delivering. http://news.ycombinator.com/item?id=1980806

How far do you plan to go with this? You could certainly make a buck or two if, say, you start offering a premium IDE with auto-complete, server-side compilation, etc. Not sure there's a big demand for this. Maybe you could sell this to enterprises and they could just deploy a lots of thin clients--save some cash on powerful workstations--and can log in from anywhere to develop and compile the code server-side and seamlessly deploy. But I'm just thinking out loud now. Best of luck.

Thanks. Except, I'm not sure Hulu qualifies as a web startup since it's a joint venture by NBC et al. I see CareerBuilder is owned by Garnett and others, too. I just watched a salesforce/chatter ad, but again, no longer startups as they are public. Just curious.

Sure, nothing against groupon.

When was the last time a web startup bought a super bowl ad slot? It seems to have been right before the bubble. Or maybe the pets.com ad was just so memorable--for the wrong reasons.

Perhaps. But grandchildren, brothers, sisters, etc can all help, not just children.

Whether it's wrong or not, it does make you wonder where Google's priorities are. How much time (cumulatively) did they spend on this and could they have spent this time improving their algorithms instead? Maybe it took a few hours, but still.

I'm glad you bring up that idea out at the end (by now, few will read it, but that's okay). It pains me to learn that some people in this great country (for cultural reasons?) don't do this. It's a "crazy" idea, but one that works very well in other countries. Latin Americans, Indians, and others will attest to that. I wouldn't call for forcing families to take care of their elderly, but I'd find ways to motivate that because it's simply sad that we have to take care of others' parents.

And yes, the whole problem with our health care system isn't so much that we're paying (we have to no matter what, either by premium increases or footing the bill for emergency patients). The problem is that it sucks at preventing major health problems that later cost more if undetected and not treated early.

Right, the cost functions could be defined in terms profit maximization (thus, the cost being lost sales, eg).

Either way, these experiments are very difficult to control because you can't fully validate models by predicting sales a posteriori with holdouts once you have the data as you don't know if some algorithm might influence certain sales not reflected in such real-life data. You could try a/b testing, but I'm not sure the Amazons are willing to run someone's models very readily. Either way, this is a good thing.

No, I'm including the whole package including health care (see 2b) and military spending which I know just about everyone agrees we need to have some of (see 4), and so on.

I should be clear: more than people who are in the middle/lower classes in terms of the effective tax rate (not just marginal), and given our current situation, it may have to be even more than it is today (along with cuts) or the alternative is that the system could collapse.

I'd like to see a graph with per capita numbers now.

On a per capita basis, California does slightly better than the US at large: about $50.8K vs. $47k US (according to the CIA/IMF):

http://en.wikipedia.org/wiki/List_of_countries_by_GDP_(PPP)_...

What a simple idea that will go a long way. There are great many algorithms/models out there that are seemingly great and do wonders on toy data, but in the real world fail not only because the data isn't comprehensive (it isn't sampled well), but also because people misuse existing models and tweak them enough to work for that data and those tweaks make it on acclaimed journals/conferences. These kinds of efforts will help tremendously to filter out all the noise because now, you'll have to show results for real-world data to argue that your algorithms can generalize and/or that your tweaks actually matter.

Besides Amazon, Google has taught us that more data is better and sometimes less complex algorithms that scale but give you only a marginal hit on accuracy are preferred; ie, cost functions matter, and that's something you don't get practice with in the academic world.

When you think about data sets to boost performance, think outside the directly dependent ones. There is so much data out there now and it's not being used. This I think will change soon.

Of all those that agree with such argument, I'd like to get an answer to these questions:

1) have you traveled to a developing|third-world country? 2) if you haven't, you should. if you have... 2a) what price tag do you put on first-class infrastructure that makes your business run smoothly, not to mention make it safer to deliver? 2b) what price tag do you pay to ensure that you have a well-nourished and healthy work-force that is more intelligent, stronger and efficient? 2c) what price tag do you pay to ensure that everyone has access to a world-class education (again, compare to a third-world country, I know ours sucks compared to others') that is again, more knowledgeable, more creative, etc and will one day help you boost margins, compete globally, etc?

And I could go on...

You see.. safety goes hand in hand with: 1) economic stability 2) health (ever wonder why some people risk their own lives to cross borders? their families are starving/dying!) 3) education 4) and much more than just a strong military, uncorrupt/just police/court system

And those things cost money. If you'd rather grow wealthy in another country, I dare you to try! And if you're already wealthy and you think taxing you more is unfair, I dare you to move to countries like Mexico where many business owners now have to pay "taxes" to drug gangs or else, they just don't go to prison, they get killed!

I'm not arguing that our tax system doesn't need reform or that current rates are fair, I'm arguing that if you want to be wealthy one day or are already, it's in your best interest to contribute more. Consider it an investment where you can. Yes, surely there is waste, but focus on that, don't whine about the fact that you're getting punished because YOU ARE NOT... you were rewarded with all these great opportunities and public goods and services you seem to take for granted, that's all (again, see my first question above).

Here's what a commenter, DogGunn, had to say (I copy below as I'm not sure it'll stay at the top):

"Article is wrong.

Mark's profile is here: http://www.facebook.com/zuck

And his like page is here: http://www.facebook.com/pages/...

Neither were hacked.

Someone had setup a fake profile at http://www.facebook.com/markzu... and fooled Techcrunch. Facebook fixed by deleting the fake profile."

A quick copy+paste pretty much confirmed it for me.

Yep. I'm sure FB will be smart about it.

What's nifty about the mechanism is that it doesn't, at least for now and according to the article, require the ever so difficult sentiment analysis. You have to explicitly "like" something or check in. Some users will have fun abusing this, but not so much more than some researchers abuse machine learning/nlp who complicate their living by putting too many constraints on a machine intelligence task. (Don't get me wrong, I love machine learning/nlp; it just doesn't have to be more complicated than it already is.)

"Apple also could use NFC to improve how it delivers mobile ads to customers’ handsets and charge higher fees for those ads, Crone said. NFC would let Apple’s iAd advertising network personalize ads to the places where a customer is spending money."

Minority-report-like world, here we come... http://www.youtube.com/watch?v=oBaiKsYUdvg

Somehow I knew this was coming and had a similar idea for a startup, except I'd make these ads more transactional oriented and give users who triggered the ad a cut, since after all they are promoting a product/service. And I'd limit it to n ads per user per week, otherwise you can get spammy behavior out of "friends" trying to maximize their credits--sound familiar?

Anyway, we'll see how that works out. I would honestly demand money if a brand tried to associate myself with them even if it makes sense. It's time to track and rewad word-of-mouth ads--consider them this now. It's clear Facebook and others are in a position to do this. You know that $50 billion+ valuation? Doesn't sound as crazy now, does it?

update: Reading some comments on mashable, I see a good counter to my point: why should FB pay you if they don't charge you to use their site? Fair enough, but at some point users deserve to be rewarded by brands/companies (not facebook) for their loyalty and for attracting them new customers.

The thesis of this lecture makes sense to me. I, however, think it's still possible to do a fair bit of great reading via blogs, Twitter (well, links shared via Twitter)and via Hacker News (take this very link, for example) and not just via great books, but it's certainly easier to get distracted by these.

Yes, you need time to concentrate on doing great things and introspecting, but it's important to open oneself up to different experiences, especially the ones that make you feel uncomfortable (but that won't hurt anyone, I would hope). That means breaking concentration and quite possibly even breaking your own moral values--or those set by your predecessors/institutions and peers, really. If not, how else will you know what's right from wrong or whether you truly dislike something or someone?

I'm not sure Deresiewicz would disagree with that, since he argues that original ideas come about when one can "make associations, draw connections, take [one] by surprise". But I thought it was worth expanding on this idea because in order to make new associations, you have to go out and experience something different that will add new neural connections.

Let's take our so called "war on terrorism". The key question I'd ask myself is, why are there such terrorists? Is it really that they hate our freedoms, our religion, etc? Every time I read some Opinion piece by an author who thinks (s)he knows why terrorists hate America, I am turned away. How the hell can they know if they haven't even been there and lived amongst them? So it saddens me that some of the so-called experts of our time are those who only read a lot and thus can articulate "their" ideas well. I don't care if you read the greatest novels and history books, if you don't get out in the field/real world also, then you won't truly understand an issue and thus, you're not an expert/(thought) leader.