Re:Patrick McKenzie (patio11) on Askolo
by scottkrager on Monday March 26, 2012 @06:32PM
(#3758961)
Lots of nuggets in that interview....
my favorite (as a competitive SEO)
"Social media: It's perfect for SEO companies which couldn't produce results in SEO, because if you fail to produce results with social media it will be utterly indistinguishable from actually producing results with social media, but if you fail to produce results with SEO anybody looking at the gross revenue graph will notice really quickly."
Re:Patrick McKenzie (patio11) on Askolo
by ced on Monday March 26, 2012 @09:59PM
(#3759591)
my understanding from undergrad and talking with people is that there is a heavy split between academic ML/AI and industrial ML/AI. One uses a lot of raw brainpower, the other uses a lot of base cunning to reduce problems that look like they require intelligence into large instances of boring math problems which were mostly solved by the 1970s.
That's a surprise to me. Can anyone here expand on that? What are the books to learn about "industrial ML/AI"?
Re:Patrick McKenzie (patio11) on Askolo
by _dps on Tuesday March 27, 2012 @01:23AM
(#3760146)
Well, I'm biased but also experienced (former academic, now industrial practitioner doing a startup). My advice to the people who think that industrial ML/AI is "just applying some base cunning to 1970s problems" is to try to trade equities and generate durable above-benchmark profits over 2-3 years (where the industrial state of the art controls hundreds of billions per year and the competition can pay $300k+/year for fresh-out-of-grad-school talent). Algorithmic equities trading is sort of the "UFC" of ML; money talks and, er, bovine byproducts walk. Having said that, I'm now out of equities trading because I figured having a startup was more than enough risk for me; so I'm somewhere between talking and walking, I guess :P.
In terms of books: I recommend grabbing as many domain-specific books as possible rather than general-purpose ML books. Look at bioinformatics, speech processing, text processing, image processing, algorithmic trading, epidemiology, system identification, adaptive filtering, etc.; each of these disciplines has its own approach to signal/feature extraction, and ML gives you a unified way to fuse multiple signals into an estimate/decision. In my experience the tricks of the trade arise from learning lots of domain-specific "hacks" and thinking about how they generalize to other problems (one example: look at the Viola-Jones feature extractor for images and think about how you might apply that in, e.g., equities trading).
Just like with programming, practical ML is best learned by just solving a bunch of problems and learning what works (informed by a theoretical framework about what can't possibly work :-)
Re:Patrick McKenzie (patio11) on Askolo
by Drbble on Monday March 26, 2012 @10:56PM
(#3759788)
Parent is talking about Agent systems and just about everything that isn't regular statistics. Basically, in the past 20 years, computers got 1000 times smarter and people didn't, so old statistical models became tractable to apply to terabytes of data, and the schools of "invent a thinking algorithm" stopped being relevant.
/slightly bitter former "academic AI" student.
It's not really Academic vs Industry, though. It is Agents and Logic vs Statistics.
The standard text is Elements of Statistical Learning. It is a grad-level and mostly theory. For goofing around in Python, Programming Collective Intelligence
Re:Patrick McKenzie (patio11) on Askolo
by ced on Monday March 26, 2012 @11:56PM
(#3759944)
I agree that Rusell and Norvig AI doesn't have much penetration yet. As for Elements of Statistical Learning...
That's the canonical textbook for ML. If industry relies on splines, boosting, and support vector machines, then it is really not that far from modern academic ML research.
Re:Patrick McKenzie (patio11) on Askolo
by dpritchett on Monday March 26, 2012 @10:17PM
(#3759648)
There was a neat tangent on this subject when one of the more recent free online ML classes came up. Basically the free course is "ML for practitioners" and the Stanford-only course is "ML for researchers". The former group is less interested in advancing the state of the art and more interested in using known ML techniques to solve business problems.
Re:Patrick McKenzie (patio11) on Askolo
by peripitea on Tuesday March 27, 2012 @01:19AM
(#3760139)
Question for the askolo people: Is this Q&A finished, or will it ever be? It isn't obvious from the UI whether there is a time window on how long Patrick will be answering questions, or whether he may continue to answer questions here indefinitely.
I ask because my normal consumption method for this sort of thing is to wait until the interview is complete to read it, but in this case I can't tell if it will actually end anytime soon.
Re:Patrick McKenzie (patio11) on Askolo
by pbiggar on Monday March 26, 2012 @09:25PM
(#3759499)
I'm astonished that someone would ask the question "marriage or YC". Surely the number of people who have somebody they'd like to marry, who would choose YC over marrying them, is negligible?
Re:Patrick McKenzie (patio11) on Askolo
by tokenadult on Monday March 26, 2012 @07:19PM
(#3759103)
you can signup without facebook too. We'll make that more obvious
I'm on board as one of the folks to receive questions, and yet every time I visit the site, it asks me to log in again, unlike almost any site I have signed up for as a regular participant. I probably do some kinds of script-blocking that only a minority of users do, so this bug in persistence of sign-in information may not have been experienced by anyone else, but I'll not for the record that I don't even see a display of a way to log in with Facebook, so that's probably a symptom of the same issue. Never assume anything about the characteristics of the user's browser unless that is mission-critical for your site. I can use other online sites (including Facebook) just fine.
AFTER EDIT: To answer your kind follow-up question, caoxuewen, I'm using Chrome with all of the following extensions installed, mostly with default settings:
Adblock Plus (Beta)1.2
Warning:
This extension failed to modify a network request because the modification conflicted with another extension.
Don't track me Google2.0
F.B. Purity - Cleans Up Facebook6.6.2
G+me for Google Plus™6.0.3
Ghostery3.0.0
Google Calendar (by Google)0.7
Google Tasks (by Google)1.0
Google Translate1.2.3.1
Google Voice (by Google)2.3.6.8
Highlight to Search1.0.36
Send from Gmail (by Google)1.12
Show Full Domain on Hacker News posts1.0
(I'll work on fixing the one conflict while you figure out what's going on with Askolo. Thanks for your help.)
Re:Patrick McKenzie (patio11) on Askolo
by staunch on Monday March 26, 2012 @06:39PM
(#3758992)
It says "Don't have Facebook?". I prefer something like "Sign up with Email". Most of the people that want to use email have FB accounts, they just don't want to use them.
Re:Patrick McKenzie (patio11) on Askolo
by patio11 on Monday March 26, 2012 @08:25PM
(#3759326)
As long as you're here: can't post from my iPad because the login link on top does nothing if I tap it. Would sincerely appreciate a fix, since light writing (like this comment and Askolo questions) are perfect for banging out at the cafe on my iPad prior to going home and doing real work on my PC.
Re:Patrick McKenzie (patio11) on Askolo
by Tichy on Monday March 26, 2012 @09:16PM
(#3759473)
The religion bit sounds a bit like the Sherlock Holmes rationale that he chose to not remember that the earth rotates around the sun because that knowledge has no practical impact on his life. So basically religion doesn't impact you, and therefore you chose to not think about it at all? But then why call yourself religious?
Maybe that way your beliefs really do no harm, except that you are a role model for others who might take it more seriously.
Re:Patrick McKenzie (patio11) on Askolo
by Tichy on Tuesday March 27, 2012 @07:09PM
(#3763917)
I read only the bit where he said that his whole world is very rational and he has only given the religion issue 3 minutes of thought. Maybe I missed something, but it seemed like a plausible explanation for somebody rational being religious. In fact to stop thinking about it after some point might be the only way to explain the existence of religion.
Why would it not be a good topic? I have often pondered asking HN how to deal with it, because I have a lot of religious friends and I am scared of losing them if I bring up religion as a topic. I would be interested how other HN users deal with that issue (assuming there are others who are not religious).
Re:Patrick McKenzie (patio11) on Askolo
by tptacek on Tuesday March 27, 2012 @07:54PM
(#3764022)
Yes, you missed the message he was trying to communicate. He was being sarcastic.
Yes, you are likely to lose friends if you insist on debating their religious beliefs as if they were favorite programming languages. You do not endear yourself to them when you refer to religion as a "hack", in either common definition of the term.
The best way not to "deal" with this problem is to avoid it. Hence: bad topic for a message board debate.
If you feel any further urge to discuss this issue with me, and aren't just trying to score message board points about religion, my contact info is in my profile and you can feel free to ping me about this. I've invited you to talk to me about religion; feel free to take me up on it. Know that your parent comment is light grey because Patrick's response was intended to communicate almost the opposite of that invitation.
my favorite (as a competitive SEO)
"Social media: It's perfect for SEO companies which couldn't produce results in SEO, because if you fail to produce results with social media it will be utterly indistinguishable from actually producing results with social media, but if you fail to produce results with SEO anybody looking at the gross revenue graph will notice really quickly."
That's a surprise to me. Can anyone here expand on that? What are the books to learn about "industrial ML/AI"?
In terms of books: I recommend grabbing as many domain-specific books as possible rather than general-purpose ML books. Look at bioinformatics, speech processing, text processing, image processing, algorithmic trading, epidemiology, system identification, adaptive filtering, etc.; each of these disciplines has its own approach to signal/feature extraction, and ML gives you a unified way to fuse multiple signals into an estimate/decision. In my experience the tricks of the trade arise from learning lots of domain-specific "hacks" and thinking about how they generalize to other problems (one example: look at the Viola-Jones feature extractor for images and think about how you might apply that in, e.g., equities trading).
Just like with programming, practical ML is best learned by just solving a bunch of problems and learning what works (informed by a theoretical framework about what can't possibly work :-)
/slightly bitter former "academic AI" student.
It's not really Academic vs Industry, though. It is Agents and Logic vs Statistics.
The standard text is Elements of Statistical Learning. It is a grad-level and mostly theory. For goofing around in Python, Programming Collective Intelligence
I ask because my normal consumption method for this sort of thing is to wait until the interview is complete to read it, but in this case I can't tell if it will actually end anytime soon.
I'm on board as one of the folks to receive questions, and yet every time I visit the site, it asks me to log in again, unlike almost any site I have signed up for as a regular participant. I probably do some kinds of script-blocking that only a minority of users do, so this bug in persistence of sign-in information may not have been experienced by anyone else, but I'll not for the record that I don't even see a display of a way to log in with Facebook, so that's probably a symptom of the same issue. Never assume anything about the characteristics of the user's browser unless that is mission-critical for your site. I can use other online sites (including Facebook) just fine.
AFTER EDIT: To answer your kind follow-up question, caoxuewen, I'm using Chrome with all of the following extensions installed, mostly with default settings:
Adblock Plus (Beta)1.2
Warning: This extension failed to modify a network request because the modification conflicted with another extension.
Don't track me Google2.0
F.B. Purity - Cleans Up Facebook6.6.2
G+me for Google Plus™6.0.3
Ghostery3.0.0
Google Calendar (by Google)0.7
Google Tasks (by Google)1.0
Google Translate1.2.3.1
Google Voice (by Google)2.3.6.8
Highlight to Search1.0.36
Send from Gmail (by Google)1.12
Show Full Domain on Hacker News posts1.0
(I'll work on fixing the one conflict while you figure out what's going on with Askolo. Thanks for your help.)
Maybe that way your beliefs really do no harm, except that you are a role model for others who might take it more seriously.
This is not a good topic for a message board debate.
Why would it not be a good topic? I have often pondered asking HN how to deal with it, because I have a lot of religious friends and I am scared of losing them if I bring up religion as a topic. I would be interested how other HN users deal with that issue (assuming there are others who are not religious).
Also, religion might be the greatest hack ever.
Yes, you are likely to lose friends if you insist on debating their religious beliefs as if they were favorite programming languages. You do not endear yourself to them when you refer to religion as a "hack", in either common definition of the term.
The best way not to "deal" with this problem is to avoid it. Hence: bad topic for a message board debate.
If you feel any further urge to discuss this issue with me, and aren't just trying to score message board points about religion, my contact info is in my profile and you can feel free to ping me about this. I've invited you to talk to me about religion; feel free to take me up on it. Know that your parent comment is light grey because Patrick's response was intended to communicate almost the opposite of that invitation.
I suppose you don't believe in sarcasm.